For most navigation apps, the shortest or fastest route is the default choice. Romanian startup Defensive Pedal is taking a different approach: helping cyclists reach their destination using the safest route instead.

Among the startups showcased in the Techsylvania Startup Alley, Defensive Pedal has developed a navigation app that analyzes road infrastructure, traffic conditions, weather, and dozens of other variables to recommend routes that minimize cycling risk rather than travel time.

“We’re essentially building Google Maps for safe cycling,” said Victor Rotariu, founder of Defensive Pedal, during an interview with Startups&TheCity.

A risk algorithm built on European cycling data

At the core of Defensive Pedal is a proprietary risk-scoring algorithm developed using European cycling accident data and extensive rider feedback.

According to Rotariu, the first versions of the model were validated with input from over 2,400 cyclists, while the current algorithm evaluates approximately 80 different factors that influence cycling safety.

Most of these variables relate to infrastructure, including:

• road type;
• roundabouts;
• tunnels;
• parks;
• road curvature;
• traffic models;
• weather conditions.

Rather than relying solely on mapping data, the platform combines multiple public datasets with its own analytical model to generate a safety score for every road segment.

According to the team, routes generated by Defensive Pedal are currently 2.1 times safer than the default cycling routes in Romania. The claim is based on an internal validation in which the startup compared 10,149 routes generated by the Defensive Pedal algorithm with Google Maps routes between the same randomly selected origin and destination pairs across Romania. The safety assessment used the company’s proprietary risk model, which incorporates cyclist risk factors identified in official European research on the causes of cycling accidents. 

AI continuously improves route recommendations

Although the core risk model is deterministic, Defensive Pedal adds an AI layer that continuously refines its recommendations.

As cyclists use the app, they can report unsafe locations, while anonymous usage data helps the system understand whether riders follow the suggested routes or choose alternatives. This feedback loop enables the algorithm to identify potentially dangerous areas that may not be evident from public data alone.

“The AI helps us continuously improve the quality of the safety ratings,” Rotariu explained.

Already live in Romania

The Android version of Defensive Pedal has been available for about two weeks and currently covers the entire Romanian road network. The iOS version also launched last week.

The startup has intentionally kept marketing efforts limited during the initial rollout, focusing instead on validating the product and collecting user feedback.

Even so, the platform has already attracted a few hundred active users from word of mouth alone, while approximately 900 cyclists have signed up for early access and are expected to join.

Helping cities improve cycling infrastructure

The technology is designed not only for individual cyclists but also for urban planners.

Defensive Pedal is collaborating with the World Bank on cycling consultancy projects, using its risk algorithm to identify dangerous streets and intersections at a highly detailed level.

Rather than simply highlighting that one city is safer than another, the platform can pinpoint exactly where infrastructure improvements would have the greatest impact on cyclist safety.

Rotariu said the company can already calculate safety scores for individual cities and plans to publish a city ranking covering most Romanian and European cities in the near future.

Expanding across Europe

While Romania serves as the startup’s launch market, the ambition is significantly broader.

The team plans to expand across the European Union within this year. Since much of the required infrastructure and traffic data is publicly available, adapting the platform to new countries is technically straightforward, although each market requires local validation, cyclist feedback, and adjustments to national traffic regulations.

A mixed business model

Defensive Pedal plans to monetize the platform through several complementary revenue streams.

The navigation app will remain free for basic users, while premium features are planned under a subscription model. Beyond consumer subscriptions, the company also sees commercial value in aggregated cycling intelligence.

Anonymous mobility insights can help outdoor brands, bicycle retailers, and local authorities better understand where cyclists travel and how infrastructure influences riding behavior, while municipalities can use the data to prioritize investments that improve cycling safety.

As cycling continues to gain momentum across European cities, Defensive Pedal is betting that safer navigation—and better data—can become essential infrastructure for both riders and the organizations building more bike-friendly urban environments.

The post Defensive Pedal wants to make cycling safer with AI-powered navigation, starting from Romania appeared first on Startups&TheCity.

Artificial intelligence dominated the agenda at the fifth edition of UNCHAIN Festival, which brought together more than 1,000 financial and technology leaders from over 40 countries in Oradea. Across two days, 170 international speakers debated topics ranging from agentic AI and digital identity to stablecoins, the Digital Euro, fraud prevention, and the future of banking.

But after speaking with executives from Visa, Finshape, Creatio, Tremend, AMICredit, and disruptive strategy training & consultant Andrew Taylor, along with other companies building financial technology across Central and Eastern Europe, one thing became clear: the most interesting conversations were no longer about AI itself.

No one questioned whether artificial intelligence would reshape financial services. That debate appears to be over.

Instead, discussions focused on a different set of questions. How do banks integrate AI into decades-old core systems? How do they ensure compliance with increasingly complex regulations? How can organizations deploy AI without compromising trust, security, or customer experience? And perhaps most importantly, what happens to employees as intelligent agents begin taking over an increasing share of routine work?

Although each interview explored a different corner of the financial ecosystem—from payments and lending to insurance, CRM platforms and digital strategy—the same themes surfaced repeatedly. Together, they paint a picture of an industry moving beyond experimentation and into a far more demanding phase: implementation.

AI is no longer the question. Implementation is.

Perhaps the strongest common thread across all eleven interviews was how much the conversation around AI has matured.

Only a year or two ago, discussions at financial conferences often revolved around whether banks should embrace generative AI at all. At UNCHAIN Festival 2026, that question had effectively disappeared. Executives no longer debated the technology’s potential. Instead, they discussed deployment models, governance frameworks, customer trust, integration with legacy infrastructure, and measurable business outcomes.

Catalin Cretu, General Manager of Visa for Romania, Bulgaria, Croatia and Slovenia, pointed to changing consumer behaviour as evidence that the market has already moved forward.

“More than 50% of online users now use AI tools to search for products and compare prices,” he explained.

For Visa, that behavioural shift is driving the development of what the company calls agentic commerce, AI-powered assistants capable of searching for products, comparing offers, and eventually completing purchases on behalf of consumers. Rather than replacing existing payment infrastructure, these AI agents rely on trusted payment networks to authenticate users, prevent fraud, and execute transactions securely. Pilot projects are already underway, with broader deployment expected over the coming year.

The discussion at Finshape reflected a similar shift, but from the perspective of traditional banking.

“Banks are no longer asking whether AI works,” said Dan Toderici, Country Manager Romania at Finshape. “They’re asking how to restructure their systems so AI can actually function.”

That distinction is significant. The competitive advantage is no longer simply adopting AI. It lies in building the organisational and technological foundations that allow AI to deliver value consistently and securely.

Anton Solodkyi, Customer Success Director at Creatio, observed the same evolution while working with banks deploying AI-driven CRM and loan origination solutions. The technical capabilities already exist, he explained. The challenge is integrating them into highly regulated environments without disrupting existing operations.

Taken together, these conversations suggest that the financial sector has entered a new phase of AI adoption. The technology itself is no longer the limiting factor. Success increasingly depends on execution—how institutions redesign processes, prepare data, train employees, and establish governance models that allow AI to operate safely at scale.

In other words, the competitive question has shifted from “Should we use AI?” to “Can our organisation actually support it?”. That challenge extends beyond technology. As Carlos Parker, Head of Arxia’s AI Acceleration Program, recently told Startups&TheCity, the future of AI inside companies may depend less on the models themselves and more on leadership, organizational culture, and change management. Through the AI Acceleration Program, Arxia focuses on helping organizations bridge exactly that gap—moving from AI experimentation to AI embedded in everyday business operations.

Legacy systems—not AI—have become the industry’s biggest bottleneck

If there was one challenge mentioned more often than any other during the interviews, it wasn’t artificial intelligence itself. It was everything that came before it.

For decades, banks and insurance companies have built technology stacks by adding new systems on top of existing ones. The result is a complex web of core banking platforms, customer databases, CRM systems and internal applications that were never designed to support AI-driven decision making.

Several interviewees described this legacy infrastructure as the single greatest obstacle to AI adoption.

Marius Scarlat, Head of Product Management CEE at Tremend, sees this challenge every day while helping banks modernize their digital services. Although AI has already been successfully deployed in projects such as customer onboarding and relationship management for major European financial institutions, including Lloyds Bank, implementation is rarely straightforward.

“Banking is one of the most regulated and traditional industries,” Scarlat explained. “Technology alone isn’t enough. You have to earn trust, understand existing processes and modernize them carefully.”

That gradual approach reflects the reality of enterprise banking. Replacing core systems is rarely an option, so most institutions are looking for ways to introduce AI without disrupting the infrastructure they already depend on.

Dan Toderici described a similar situation from Finshape’s perspective. Rather than replacing legacy platforms, the company builds a middleware layer that sits between existing banking systems and modern AI-powered applications. Its role is to organize, clean and contextualize fragmented customer data before it reaches AI models.

Without that intermediate layer, even the most advanced AI solutions struggle to produce reliable results.

Anton Solodkyi, Customer Success Director at Creatio, believes this explains why many AI initiatives fail to move beyond the proof-of-concept stage.

“Organizations often focus on the AI itself,” he said. “But the bigger challenge is understanding and documenting the business process you’re trying to automate. If the process isn’t clearly defined, an AI agent won’t fix it.”

His observation highlights a recurring misconception surrounding AI adoption. Automation doesn’t compensate for operational complexity—it amplifies it. Organizations with fragmented processes simply expose those weaknesses faster when AI enters the equation.

The same challenge extends beyond banking.

Sándor Székely, who develops embedded insurance solutions in Hungary, said that launching new AI-powered insurance products is rarely limited by the technology itself. Instead, the biggest delays occur inside insurance companies that must integrate new digital products into legacy policy administration systems built years—or even decades—ago.

The technology is ready.

The infrastructure often isn’t.

Across banking, insurance and enterprise software, the pattern was remarkably consistent: AI is exposing technical debt that many financial institutions have accumulated for years. Before organizations can become AI-first, they first need to modernize the digital foundations on which AI depends.

In the AI era, clean data has become a competitive advantage

Closely connected to legacy infrastructure is another theme that surfaced repeatedly throughout the interviews: data quality.

Every executive agreed that AI is only as good as the information it receives.

That may sound obvious, but in financial services it represents one of the industry’s biggest transformation challenges.

Banks typically operate dozens of disconnected systems built over many years. Customer information is scattered across different databases, product lines and business units, making it difficult to create the unified view that modern AI applications require.

“Banks are investing heavily in AI,” Dan Toderici explained, “but before AI can generate meaningful insights, the data has to be structured, cleaned and connected.”

Finshape addresses this challenge through a customer data platform that creates a unified customer profile across multiple banking systems. Once that foundation exists, AI can deliver personalized financial recommendations, targeted marketing campaigns and intelligent customer interactions.

One of the company’s flagship examples is its collaboration with Raiffeisen Bank Romania. What began as a digital loyalty application has evolved into an AI-powered personalization platform with more than one million active users. The system analyzes customer behavior and transaction history to deliver highly personalized offers inside the bank’s mobile application—demonstrating that AI creates value only after the underlying data has been properly organized.

Creatio faces similar challenges from the CRM side.

According to Anton Solodkyi, many organizations underestimate how much preparation is required before AI agents can automate business workflows.

“Structured information is the foundation,” he explained. “Without it, even the smartest AI agent has nothing reliable to work with.”

Tremend has reached the same conclusion internally. Beyond client projects, the company is increasingly using AI to accelerate software development and modernize legacy code through proprietary tools such as Backlog AI and Slingshot. Even there, however, human review remains essential to validate AI-generated outputs before they become production-ready.

Taken together, these conversations suggest that data has become the new competitive advantage in financial services.

The race is no longer about who has access to the latest language model. Those models are becoming widely available.

The real differentiator is who has built the cleanest data architecture—and therefore who can generate the most reliable, secure and context-aware AI experiences.

AI is changing work, not eliminating it

One of the clearest divergences between public perception and real-world deployment of AI in financial services is what actually happens inside organizations once automation is introduced.

The expectation is often substitution. The reality, across multiple interviews, is reconfiguration.

At AMICredit, the introduction of AI agents fundamentally changed how work is distributed inside the company—but not in the way many might expect. Rather than reducing headcount, the organization expanded.

Karoly Totos, CEO of AMICredit, described a system where AI handles repetitive, structured tasks such as onboarding and customer communication, while human employees shift toward higher-value operations.

“We’re hiring more people, not fewer,” he explained. “AI agents handle repetitive tasks, making the whole organization more productive.”

The impact is measurable. Onboarding workflows that previously consumed up to 40% of employee time have been largely automated. IT costs have been reduced by nearly half, while customer retention in AI-assisted flows has increased by 15–20%.

But the deeper change is not operational efficiency. It is the redefinition of what human work inside financial organizations actually is.

A similar shift is visible in broader discussions about the labor structure of financial services.

Visa’s perspective on agentic commerce highlights a future where AI systems increasingly handle discovery and comparison tasks on behalf of users. Andrew Taylor, who studies strategic disruption, frames this shift in broader terms: routine cognitive tasks are gradually moving toward machine execution, while humans transition toward supervision, validation, and system-level decision-making.

Across all interviews, a consistent pattern emerges: AI is not removing humans from the system—it is moving them up a layer in the decision hierarchy.

Human oversight remains a structural requirement, not a temporary phase

Despite rapid advances in automation, no interview suggested a future in which financial systems operate without human intervention.

Instead, a consistent governance model appears across banking, insurance, CRM systems, and payment infrastructure: AI proposes, humans decide.

At Creatio, this principle is embedded directly into CRM and workflow automation systems. AI agents assist in loan origination and customer service processes, but final decisions remain with human operators.

Anton Solodkyi described this as a design constraint rather than a limitation.

“AI generates recommendations, but humans make the final decision,” he said.

The reason is not technological weakness but regulatory and operational reality. In highly regulated industries such as banking, full autonomy introduces unacceptable levels of risk—both in terms of compliance and reputational exposure.

Tremend’s experience reinforces this view. Even in advanced implementations such as AI-driven onboarding and relationship management systems used by major European banks, human validation remains a core component of the workflow.

Visa’s experiments with AI-driven commerce also depend on human confirmation before transactions are executed, particularly in early deployment phases.

Across all cases, the same architectural principle appears: AI systems are being designed as decision-support layers, not decision-making authorities.

Regulation is no longer a barrier—it is a design constraint

If earlier phases of fintech innovation often framed regulation as friction, the interviews suggest a more mature relationship has emerged between AI systems and regulatory frameworks.

Rather than slowing adoption, regulation is increasingly shaping how AI systems are built from the ground up.

In insurance, compliance requirements were a central factor in deployment timelines. Sándor Székely noted that achieving regulatory approval for AI-driven insurance products required extensive validation processes, including data privacy assessments and alignment with national authorities.

Even when the underlying technology is ready, institutional approval cycles can extend implementation timelines to several months.

Similarly, Finshape’s work with banks across Central and Eastern Europe shows that GDPR compliance and data governance are no longer treated as external constraints but as embedded design parameters in system architecture.

Narcis Nagy highlighted a related dynamic: organizations are under pressure to demonstrate AI capability quickly, but must still align with strict European regulatory frameworks such as the EU AI Act.

This creates a structural tension between speed and compliance—but also a clear directional shift. Instead of resisting regulation, companies are increasingly designing systems that assume it from the start.

The result is a quieter but significant transformation: regulation is becoming part of the AI architecture itself.

The pressure to adopt AI is driven by competition, not strategy

While much of the discussion around AI adoption focuses on technological readiness, one of the more subtle insights from UNCHAIN 2026 is that the primary driver of adoption is not internal transformation agendas.

It is external pressure.

Organizations are not adopting AI because they have fully mapped out long-term strategies. They are adopting it because competitors already have.

Narcis Nagy, General Manager & Cofondator at RELOAD CONCEPT described this dynamic directly, noting that the pressure to implement AI often comes from the market rather than internal leadership teams.

“The pressure comes from competition,” he explained. “Not from within the company itself.”

This creates a paradox. Companies accelerate AI initiatives not necessarily because their systems are ready, but because standing still is no longer an option.

Andrew Taylor’s strategic framing adds another layer to this observation. In environments where technological change is accelerating, organizations increasingly shift from predictive planning to adaptive response. Strategy becomes less about forecasting outcomes and more about reacting to systemic shifts initiated elsewhere in the ecosystem.

Across both perspectives, the implication is clear: AI adoption is becoming less of a strategic choice and more of a competitive necessity.

Conclusion: finance is being rebuilt, not disrupted

Across the eight interviews spanning payments, banking infrastructure, insurance, CRM systems, and strategic consulting, a coherent picture emerges.

Financial services are not undergoing simple technological disruption. They are undergoing structural reconstruction.

Artificial intelligence is already embedded across multiple layers of the financial system. But its integration is constrained by legacy infrastructure, fragmented data, regulatory frameworks, and institutional trust.

As a result, the dominant model emerging across the industry is not autonomous AI-driven finance, but supervised intelligence operating within tightly governed systems.

Visa’s agentic commerce, AMICredit’s AI-driven onboarding, Finshape’s data infrastructure, Tremend’s enterprise deployments, and Creatio’s CRM agents all point toward the same reality: AI is becoming embedded, but not independent.

And perhaps the most important insight from UNCHAIN 2026 is this:

The defining challenge of the next phase of financial transformation is not building more capable AI systems.

It is building financial institutions capable of absorbing them.

Read also: Riding the AI Wave: Why the Real Challenge Is Not Technology, but Change, a related interview on how organizations can navigate the shift from AI experimentation to execution.

The post Beyond the AI Hype: What Fintech Leaders Revealed About the Future of Financial Services at UNCHAIN Festival 2026 appeared first on Startups&TheCity.

AI is no longer a futuristic concept—it is a global tsunami hitting every industry from the United States to Chile, Uganda, and Romania. Yet, while executives rush to adopt AI just to keep up with the competition, the real battle is being fought on the old battlegrounds of digital transformation: people first, processes second, technology third.

This contrast between market rush and operational reality was the catalyst for a recent workshop I’ve attended, led by Carlos Parker, head of Arxia’s AI Acceleration Program, in Cluj-Napoca, Romania. Working with a local consultancy firm, the discussion proved that while AI models are evolving faster than ever, the real art lies in learning how to surf the wave rather than fighting it.

Carlos, who has delivered workshops across Europe, Latin America and Africa, says the conversation around AI has become universal. From Romania and Ukraine to Chile and Uganda, businesses everywhere are talking about artificial intelligence. Yet, according to him, the technological challenge itself is not the hardest part. 

“The real issue is not the technology itself,” he explains. “It’s about operationalizing it. How do companies integrate AI into their workflows? How do they measure return on investment? How do they help employees adapt?”

In his view, AI is not fundamentally changing the classic rules of digital transformation. The same principles still apply: people first, processes second, technology third.

AI Is an Accelerator, Not the Core

One of the strongest ideas emerging from my discussion with Carlos was that AI should not be treated as a goal in itself.

“The core has always been process optimization,” Carlos says.“You spend far more time defining workflows, defining policies, and managing organizational change than teaching people not just how to prompt, but also how to integrate and orchestrate AI tools effectively.”

This perspective contrasts sharply with the current market trend, where many organizations rush to “adopt AI” without clearly defining why they need it in the first place.

According to him, many executives are under pressure to demonstrate AI adoption simply because competitors are doing the same. Some companies already demand internal AI KPIs, even when employees are using AI for tasks that provide little practical value.

“The fact that you can use AI doesn’t mean you should,” he argues. “The important question is: does it actually improve efficiency or solve a real problem?”

Surfing the Wave Instead of Fighting It

Asked whether people should fear AI, Carlos used a simple metaphor.

“AI is like a wave,” he says. “You can try to fight it, or you can surf it.”

He believes the acceleration of AI development is unlikely to slow down. Governments, corporations and entire regions now see artificial intelligence as a strategic race tied to competitiveness, economic power and geopolitical influence.

“The United States, China and Europe all understand this is a priority,” he explains. “Once countries start investing resources at this scale, it becomes very difficult to stop.”

Competition itself becomes the engine. If one country or company believes another is advancing faster, everyone feels pressured to continue investing.

Beyond Automation: The Bigger Philosophical Question

The conversation also moved far beyond productivity and automation.

Carlos believes society is underestimating the deeper implications of AI — especially if artificial general intelligence (AGI) eventually becomes reality.

“If AI becomes better than humans across all economic domains, then we enter a completely different scenario,” he says. “And we don’t really have political or economic frameworks prepared for that.”

Unlike previous industrial revolutions, which created new categories of jobs alongside automation, AGI raises a more difficult question: what happens if machines can improve, maintain and optimize themselves?

Modern economic systems — whether capitalist or socialist — are fundamentally built around human labor. If work itself is no longer necessary, entire ideological structures may need to be reconsidered.

“How much meaning in our lives is connected to the work we do?” he asks. “What happens if that disappears?”

Eastern Europe’s Unexpected Advantage

One particularly interesting part of the discussion focused on Eastern Europe and its relationship with technology adoption.

While AI adoption among businesses remains cautious, Carlos sees the region as having a major long-term advantage due to its strong engineering and technical culture.

Countries such as Romania, Ukraine, Estonia and Moldova inherited a deep emphasis on science and engineering education — something he believes still shapes society today.

“In Eastern Europe, engineering and technology are culturally valued,” he says. “You see it in the number of programmers, in technical education, and even in the role of women in engineering compared to many Western countries.”

Romania, in particular, was mentioned as a country that historically adapted quickly to new technologies. From internet infrastructure to software development talent, the region may ultimately prove more flexible than some Western European economies traditionally perceived as more advanced.

At the same time, resistance to change remains universal.

“People adopt technology slowly everywhere,” he notes. “A lot of it is cultural. Change always involves risk.”

AI Adoption Is Ultimately About Trust

One of the workshop’s recurring themes was that companies are often less afraid of AI itself and more afraid of uncertainty.

Managers understand that change is coming, but many do not know where to begin.

That is why, Carlos argues, successful AI implementation should resemble a gradual organizational transformation rather than a simple software rollout.

“You don’t just buy licenses for everyone and expect magic to happen,” he explains. “Successful AI adoption requires a strategic, company-wide transformation process implemented gradually. You start with focused teams or use cases, validate what actually works in practice, measure results, build internal know-how, and then scale adoption step by step across the organization.”

In other words, the future of AI inside companies may depend less on the models themselves and more on leadership, organizational culture and change management.

The technology is already here. The harder question is whether people, businesses and institutions are ready to evolve with it.

And this is precisely the gap that Carlos aims to address through Arxia’s AI Acceleration Program — a framework designed to help companies move beyond experimentation and actually operationalize AI inside their organizations.

The post Riding the AI Wave: Why the Real Challenge Is Not Technology, but Change appeared first on Startups&TheCity.

The inaugural Cluj Hackathon, the first hackathon in Romania focused on Artificial Intelligence, and Generative AI took place between June 16-18, 2023, at the Bosch Engineering Center Cluj. More than 115 participants, organized into 43 teams, enthusiastically joined the hackathon, making it one of the most successful events of this kind in Romania.

For three days, the participants had the opportunity to build and develop revolutionary applications and platforms, with access to high-quality mentoring, a wide range of resources, and networking opportunities, with participants from across the country. 

Remarkably, the youngest contestants in the hackathon were two boys, aged only 12, who independently acquired coding skills at home. Motivated to learn new things and connect with like-minded individuals, they opted to participate. With the guidance of mentors, they successfully developed an AI story generator. “It feels immensely gratifying when you code something and witness its functionality. That’s what drives me,” expressed Prisăcaru Thomas, one of the youngest participants in the hackathon.

The inception of the hackathon took shape when Adrian Vlasiu discovered the boundless potential of ChatGPT. His early engagement with OpenAI’s groundbreaking technology evoked a “deja-vu” moment, similar to the iPhone’s initial market launch. Fuelled by this profound insight, Vlasiu resolved to orchestrate a hackathon, thereby providing a platform for others to explore and harness the transformative power of this remarkable technological advancement.

“I wanted to organize the biggest hackathon in Cluj. And I wanted to organize the Untold Festival of hackathons to set a new standard in terms of satisfaction for the participants, satisfaction for the sponsors, and the whole overall experience of the competition. The engaging format of the Cluj Hackathon proved to be a winning recipe because the participants saw a different hackathon format. It’s not just about the competition or the prizes; it’s all about building community, engaging with each other, and succeeding by helping others succeed. We had very generous sponsors and top-class companies like Bosch that enabled us the hackathon location within Cluj-Napoca at the Bosch Engineering Center Cluj, one of the smartest buildings in the office scene in Romania. Untold Festival and Banca Transilvania supported our initiative, giving us a huge boost of confidence. And, of course, we had a scorching topic, artificial intelligence, and ChatGPT.”, said Adrian Vlasiu, founder of the Cluj Hackathon.

Out of the 43 teams which took part, team TestBusters emerged triumphant as they won the first prize in the hackathon. Their creation, an AI-powered solution for automated software testing, showcased ingenuity that pleased all the judges. There were other outstanding projects, and eventually, the score was tight. However, when measured against the hackathon’s primary objective of delivering a functional product within the allotted time, TestBusters stood firmly at the forefront, embodying the essence of the event.

“We have created a solution that uses artificial intelligence to test applications automatically, but not only does the test, it also creates tests automatically. With artificial intelligence, it’s possible to automate more and do more. Basically, those tests are needed in every software or manufacturing firm, so we can solve them by automating them. The goal of this kind of application is ease of use. The product manager or the product owner can use it because it’s very simplified; you don’t need to know how to code; you need to fill out a form, and the computer will be able to understand what you want to do.” explained their solution, Gaëtan de Smet and Stan Ciprian, two of the members of the winning team.

The second place was awarded to the Tech Titans, who worked on detecting anomalies in a dataset from a sensor system (car batteries). Their project underscores the importance of data analysis in solving real-world issues. And, not to be overshadowed, Brainiax took the third spot for their effective solution for detecting potholes on the road. As one of the lead engineers from Bosch pointed out, detecting and analyzing potholes on the road in real-time is one of the toughest engineering challenges in the auto industry today.

A team of mentors from diverse backgrounds in technology and entrepreneurship extended unwavering support to the hackathon participants. Their invaluable expertise and guidance catalyzed innovation and critical thinking, nurturing an environment where teams could thrive and unleash their true potential. Through their mentorship, these seasoned professionals empowered the participants, enabling them to transcend boundaries and realize their ideas.

“We started with 43 teams. I was amazed by how motivated they were and how fast they developed a functional prototype that fit the purpose they created it for. It was an excellent start for the participants, and I think that maybe all the startups should go through a hackathon at some point to kickstart their idea.”, said Smaranda Ignat, mentor.

“I’ve seen some great ideas, for example, one team taking care of the blind people and trying to integrate them in society by offering an easier integration in the digital world, like websites that can be translated into Braille, using speech-to-text. I’ve learned a lot from the teams. I believe this mentorship experience happened from both sides. I hope to see at least one or two of the teams hopping to the next stage and talking to some investors.” added Victor Dinică, mentor.

The host for the hackathon was none other than the Bosch Engineering Center Cluj. Renowned for its cutting-edge work in developing software and hardware solutions for automated, electric, and connected driving technologies, the center boasts an end-to-end approach. Bosch Engineering Center Cluj is a pivotal software hub in the realm of automated driving innovation, from crafting the software integrated within the vehicle to the cloud-based software that facilitates seamless connectivity.

“We’ve hosted the Cluj Hackathon to enable the community to grow because AI is a promising technology. Reflecting on our collaborative internal working environment, one of our goals as an engineering center is to develop with the local and international communities. We cannot do a one-man-show these days, it takes teams to build big things, and as a reflection of that, we want to be part of the community, the ecosystem and develop together with the universities, other companies, and startups.”, said Cătălin Golban, responsible for the Department developing Computer Vision technology for automated driving at Bosch Engineering Center Cluj.

When considering productivity enhancements, implementing AI can yield significant optimizations across various job domains, encompassing software development and testing. However, the realm of technology development is also profoundly impacted. Notably, Bosch has been harnessing the power of AI for an extensive duration, commencing a decade ago with the ambitious objective of assembling a team specializing in computer vision for automated driving. The outcomes of this endeavor have been truly remarkable, with millions of cars worldwide currently equipped with AI-based technology crafted by the Engineering Center situated in Cluj-Napoca.

“For us, AI is not new, but we see more and more new models emerging, which help us solve problems that previously were either unsolvable or maybe were solved by rather traditional algorithmic methods. The ecosystem consisting of startups and companies needs to adapt, it needs to professionally convert people in the software area to get familiar with using AI when developing features and as productivity tools. I think we should, as a community, not miss that train.

I also think such hackathons’ social and learning parts are essential. People came here to learn not only about technology but also how to pitch ideas and present what they think. I believe that big things can be done in teams, through collaboration, by sharing know-how and exchanging ideas; this is how the community will grow. Events like Cluj Hackathon are a big boost for the community.”, added Cătălin Golban.

With their sights set on the future, the organizers expressed their aspiration for the Cluj Hackathon to become an enduring tradition, extending its legacy into the upcoming year. Driven by an unwavering commitment to excellence, they seek to elevate the standards of hackathon organization, aiming to amplify the event’s quality and the unparalleled value it delivers to participants. By striving for the maximum return on investment, the organizers are relentlessly creating an even more remarkable and rewarding experience for all involved.

The post The story behind the success of Cluj’s biggest hackathon: a journey of community and innovation appeared first on Startups&TheCity.

Veridion (formerly known as Soleadify) announces a $6 million late seed round to expand to US and further develop their proprietary AIs that are specifically trained to understand the global business landscape, enabling their clients, including tier 1 consultancies and market intelligence firms, to access reliable data on 80 million businesses worldwide, saving thousands of hours of manual research and achieving record accuracy and refresh rates.

The deep tech startups raised the seed round from LAUNCHub Ventures, OTB Ventures, Underline Ventures, and existing investors Day One Capital and GapMinder Venture Partners.

Veridion leverages the power of AI to revolutionize big data and to comprehend billions of webpages every week in order to build and maintain a highly accurate, global scale, near real-time, single-source-of-truth database of company data made available through a Complex Search API and a Matching & Enriching API.

They are one of the first companies to deploy AI at this scale for data acquisition, setting them apart from the traditional firmographic data players that buy data from government registries as well as simple web-scraping alternative data providers.

Firmographic data is information that can be used to categorize organizations by a host of information such as business activity, products or services offered, location, employees, industry, technology offered etc. Enterprises are missing reliable, up-to-date information about market participants and a lack of those crucial insights leads to sub-optimal decisions and challenges when navigating through dynamically changing environments.

Veridion is positioned to be a firmographic intelligence leader, boasting a database of over 60 data points on over 80 million businesses with record accuracy rates and the highest refresh rate on the market.

Veridion saw solid growth figures in 2022, with over 3x increase in recurring revenue and an impressive zero churn in clients. Their biggest clients include, among others, tier 1 consultancies with specialized procurement and insurance teams, one of the largest global insurtech platforms, procurement platforms, and top tier market intelligence companies. There are multiple reasons for this, key among which is their highly granular data, the control their APIs offer on confidence scores and their internal matching engine which allows them to pinpoint businesses at unprecedented rates.

The name change is intended to highlight the company’s unique standing in the market, their commitment to being a single source of truth and reflect the company’s ambitious plans for future growth.

Florin Tufan, founder and CEO, Veridion: “We believe that the next generation of business decision engines, regardless of industry or vertical will require a completely new format of data. Predictive models will rely on AI more than human intuition and will be capable of processing exponentially more data as their core resource. For this, the data needs to be fresh, granular, easy to work with, and integrate-able across a wide array of processes. This has been our focus from the beginning and with this round, we’ll be able to double down on our approach.”

The firmographic data market in which Veridion operates is also known as the business information market and was valued at $211 billion in 2022 and is projected to grow towards $291 billion by 2027, with new technologies like AI representing a game-changer for big data. According to Gartner’s Data Quality Market Survey, the average financial impact of bad data on organizations in 2022 was estimated to be around  $12.9 million per year.

Recent developments in ML capabilities unlocked the doors for Veridion to be positioned quite differently from other market participants. Players such as Delphai, Planck Data, and Grata have collectively raised over $110M in the past 7 years of operations and development. While companies like these cover local markets and company records in the 10 – 20 million range, Veridion boasted over 70 million company records within their first two years of activity.

Delphai covers primarily private markets for M&A, prospecting and supplier sourcing, segments where Veridion brings in-depth data on hundreds of millions of products and services and business activities in all commonly used taxonomies worldwide such as NAICS, NACE, SIC, etc on companies from all around the world. Planck Data is focused on commercial underwriting but only covers the US market as opposed to Veridion’s global coverage and Grata, which is focused primarily on deal sourcing also lacks the in-depth perspective that Veridion brings on specific company descriptions, risk signals and other rapidly expanding data points that Veridion extracts on the companies they oversee.

The biggest name in firmographic data is Dun & Bradstreet, with over 200 years experience in the field. Dun & Bradstreet and Veridion may serve slightly different market segments, but their differences are pronounced when it comes to data products and capabilities. Veridion offers a distinct advantage in its coverage of the SMB landscape, with regularly refreshed data and in-depth product-level information and retail locations. The NLP-based approach employed by Veridion also enables the extraction of unique data attributes for specific industries, such as specific underwriting and supplier risks, enabling the company to better meet the needs of its clients through custom data solutions.

After validating their data with large enterprise clients in commercial insurance underwriting, supplier discovery, and private market analysis, Veridion is releasing self-service APIs to enable faster data integration for potential clients and open up to new markets in data-centric SaaS or other early-stage solutions.

Veridion plans to continue building its US footprint by opening a New York office in 2023 to more closely support their growing base of clients in the region.

The post Bucharest-based Veridion secures $6 million to further develop their AIs tracking 80M businesses appeared first on Startups&TheCity.

XVision, a medical platform that analyzes lung X-rays using artificial intelligence, started to be used for the identification of the patients with lung injuries, associated with COVID-19. The solution, offered pro-bono, became part of the mobile laboratories for the detection of coronavirus, located in the outdoor spaces of the Institute of Pneumophysiology “Marius Nasta” in Bucharest, respectively of the Emergency Reception Units of the Emergency Clinical County Hospital in Timisoara.

This new way of medical triage will reduce the risk of contamination of other patients already admitted, but also of the medical staff. In addition, this MedTech solution supports the medical team in managing the extremely high workload during this period.

Thus, through a partnership facilitated by the Institute of Pneumophysiology “Marius Nasta” in Bucharest, XVision is used to support the triage of patients showing signs of infection with the SARS-COV-2 virus. This hospital unit has two mobile laboratories equipped with X-ray devices for tuberculosis detection. One of the laboratories was sent to the Timișoara County Hospital to facilitate the patients triage process. These mobile units will be used until the end of the pandemic to prioritize patients who require lung X-rays.

The triage system is based on an evaluation score developed by XVision’s radiologists’ team, benchmarked with the scores and standards used globally, at the time being. Based on the algorithm, it describes the similarities between the radiological patterns in the patient’s X-ray and those found in COVID-19 cases.

“XVision is a system that can be replicated and adapted in other hospitals in Romania, too, being a real support for radiologists in managing the large volume of medical images and at the same time reducing the degree of infection of the medical staff. We are already discussing with a few hospitals that we could support during this period. We express our desire to help pro-bono as many as possible medical institutions that receive patients infected with this virus”, says Stefan Iarca, one of the co-founders of the XVision project.

As part of the triage, after the X-ray is done in the mobile lab, the XVision application provides a very quick result that is analyzed by the doctors, along with the initial X-ray. Based on it, a decision is made regarding the respective medical case. XVision detects signs of pneumonia, even in incipient form, and detects the degree of resemblance to the radiological pattern of pneumonia with that specific to COVID-19.

“For the County Clinical Emergency Hospital << Pius Brinzeu >> from Timișoara, the collaboration with the Marius Nasta Pneumophysiology Institute in Bucharest is a welcome and helpful one for both the medical staff and the patients who arrive in the Emergency Receiving Unit within the healthcare unit. The engineers from XVision, with whom the hospital has already implemented an artificial intelligence program in the Radiology Laboratory, have the opportunity to test this platform, which has proven to be useful for the imaging diagnoses investigated within the Radiology Section, as well as within the mobile caravan, which arrived with us at the end of last week.

Our goal is to minimize the risk of contamination for both patients and medical personnel who are in the front line in the battle with Covid-19, beyond all the truly rigorous procedures we have set up and the protective equipment for all the medical staff and for the patients.

The program developed by XVision experts facilitates the work of radiologists, by accurately detecting lung lesions specific to the new coronavirus. Images provided by this software are viewed and checked in real time so that the diagnostic errors are minimal” stated Asist. Univ. Dr. Raul Pătrașcu, Manager of the County Clinical Emergency Hospital “Pius Brinzeu” from Timișoara.

More details on how XVision platform can be used to evaluate X-rays of patients suspected of COVID-19 are available here.

The application developed by XVision, is the first of its kind in Romania. It was implemented initially in the Timișoara County Hospital where it helps doctors analyze up to 200 lung X-rays per day, meaning 20% more than in the past.

Currently, the application generates data related exclusively to the chest area using digital X-rays as basic information. Currently the team is working on developing new algorithms that will be able to analyze chest and cranial CTs.

XVision is a Romanian startup founded in 2018 by a multidisciplinary team from Timișoara, passionate about artificial intelligence, deep learning and optimization of medical processes.

XVision uses artificial intelligence algorithms for lung X-rays analysis, assisting physicians in the diagnostic process. With the help of XVision platform, medical imagery benefits from the technological advantage offered by artificial intelligence ensuring a more efficient medical act, centered on the quality of the patient-doctor relationship.

The post XVision scans patients suspected of being infected with COVID-19 in mobile laboratories in Bucharest and Timisoara appeared first on Startups&TheCity.

In the context of the COVID-19 pandemic, Timisoara-based startup XVision has announced it offers free access to its AI-based healthcare platform to any doctor who believes that this platform could ease his work in the coming period. The platform developed by XVision allows faster and easier analysis of chest radiographs with 20% more X-rays being interpreted daily, the application signaling those with pathologies and prioritizing them for visualization by radiologists.

XVision observed that a record number of chest radiographs were recorded in a partner hospital, the increasing workflow leading to the exhaustion of the doctors and to system overload. XVision uses artificial intelligence algorithms for the analysis of the pulmonary X-rays, among the radiological aspects identified being those directly correlated with pneumonia, a pathology that is also caused by COVID-19.

The doctors who would like to access the platform should contact XVision here for more details.

The startup was established two years ago, had an accelerated growth in 2019, reaching to expand nationally, the platform now being tested by hospitals in Timisoara, Oradea and Bucharest.

Last year XVision was selected in the Techcelerator acceleration program, through which they raised a total investment of € 100,000 from GapMinder Venture Partners. Techcelerator has recently launched an initiative intended to bring in one place Romanian startups that could ease the crisis situation generated by the COVID-19 pandemic. Startups can register here.

The post Romanian startup XVision offers doctors its AI analysis system to help diagnose COVID-19 faster appeared first on Startups&TheCity.

Cluj-based teleportHQ announced that it has launched the beta version of their AI-powered platform for automated generation of front-end code and design, a solution meant to keep up with the huge unmet global demand of experts in building user interfaces. The platform is addressing developers by automating their front-end work and learning their way of design, their project, and their industry scope. The project is the result of a sustained R&D effort of more than 3 years, partly financed by the European Union. 

In a digital world where everything is going to be displayed on screens, there is an increasing demand of experts in building user interfaces, and there’s simply not enough front-end developers to keep up with the demand. Today, there are about 25 million developers  around the world, from whom about 5-6 millions are front-end developers. There is a huge need for automation in the front-end development field in order to reduce costs and respond to the demand. teleportHQ decided to respond to this challenge by inserting machine learning in their platform.

When a user starts with an idea of making a website, they would have something that you would call “low definition information.” It is an idea or maybe a wireframe, and in order to get the final code from the wireframe there are a lot of decisions that have to be made. Usually those decisions are made by a designer then integrated by a developer, and with your intention and their knowledge and work, finally you get the website. The automation part of this process is all the knowledge that a team of experienced designers and developers from teleportHQ have put together and stored in what they internally are calling ‘processors’. One after the other, those processors take decisions about how your wireframe should be transformed into the final code.

Learning from their extended experience in front-end development, the teleportHQ team has found that many of the decisions that designers and developers do are similar; there are patterns. They’ve started to automate everything that they’ve detected as patterns, so that when a user will create a wireframe, the algorithm will “understand” the user intention, and will generate for them what should be the final result and code. 

teleportHQ is now working on an engine that creates predictions, which means while you’re working, the platform will propose to you the best next steps according to the training that it has received about your own behavior in the platform. This leads to a hyper personalized experience per user.

teleportHQ’s platform consists of an ecosystem composed of several parts, the most visible and understandable part being what they are calling ‘the Playground’, a SaaS platform where users can build user interfaces in a visual and efficient way.

The Playground has a wizard-like experience, what you see is what you get, but it is really advanced, meant for professional use. teleportHQ plans to add features to the platform, new experiences that are oriented for designers, content providers, and marketers. In the end, even business owners will be able to use the platform and collaborate with their peers and colleagues.

In the Playground background, a set of machine learning engines would kick in. Some of them make up the detection and transform drawings or wire frames directly into code. Other engines would make code optimizations and intermediary decisions or predictions about the user’s next actions. All these components are working together in one single ecosystem and are dramatically accelerating the speed of producing front-end development or websites.

The teleportHQ ecosystem is based on an intermediary representation format, called UIDL (an acronym for User Interface Definition Language), which could very likely become an open user interface definition language in the future. It is a JSON data structure that is code agnostic. Open source code generators are able to read this data structure and transform it into whatever type of code, supporting almost 10 frameworks to this day. On top of those two elements, the UIDL and that code generators, teleportHQ has built their platform, where you can visually do all the front-end design, with different collaboration capabilities.

Building a user interface is quite a complex process, involving up to five different types of roles. This can include the stakeholder who wants something to be done, the product owner – a proxy for the business owner, the designer, the developer, and finally the customer who would close the feedback loop. 

The discussions between all these users are usually quite hard because they have different languages and different goals. One of the most frequent issues is the interaction between designers and developers who have radically different ways of thinking. Designers think in what you would call an absolute positioning way and the developers think  more in how their code should be implemented. It is a huge pain, and a solution is yet to be found for the seamless transition from design to code or from an idea/wireframe to code. 

Most big companies in today’s industry, especially in the design side, consider that the way to fill this gap is to add some extensions to their design platform, in order for them to generate the code. teleportHQ believes that it should be done the other way around so they are building the bridge from the developer side towards the designers. This is the reason why they have launched this stage of the platform with the developers in the forefront of their minds. In the next iterations, they’re going to build layers that are more appealing or easier to use for designers, then add new layers that are more appealing for content owners, then features for marketeers, and finally features specifically designed for stakeholders. In the end, you need a context in which all those roles and people would meet together in order to finalize and very quickly discuss the final product. And this is the goal that teleportHQ ultimately wants; to have a completely integrated collaborative platform in which all these roles can see at any time the status of the project and if anybody is changing any part of the project, everybody else is kept in the loop.

The platform is professional-oriented and made for production, so this is the reason why they call it low-code, in the sense that is not a no-code platform (where you wouldn’t need any code knowledge). A minimum understanding of HTML and CSS will be required to use the platform, so it is a productivity tool, not just a consumer tool.

The world is going to have a very large population of digital citizens such as developers or students that want to learn front-end design, designers that want to learn to code or make a prototype, and even marketeers that want to make simple static websites. For this kind of users, teleportHQ offers a classical SaaS model in which you will need a bit of coding knowledge but you won’t have to type all the code. It will be a local platform that has hosting included.

But the most interesting cases are in the B2B world, where teleportHQ already has discussions and is doing prototypes with several leaders in the banking industry, the automotive industry, and several other industries, even in cloud hosting. As a result of using teleportHQ, those companies could be able to prototype their interfaces 10 to 50 times faster than they do today.

teleportHQ is also in discussion with VR partners that are interested in having the 2D experience that they’re building but in 3D, where instead of making the wireframes by using the mouse or taking pitches from the whiteboard, you would just draw them into the air while you are in a conference and maybe in a very close future just directly with holograms so that you don’t even need glasses or other augmenting parts. The first VR prototypes will start next year, and the first demos will be available online.

During their beta-launch event in Cluj, Paul Brie, teleportHQ’s CEO, also announced that they have started an important strategic partnership with the Katholieke Universiteit Leuven, which has one of the main experts in the front-end development industry and will start making a technological transfer of their past 10 years of R&D towards teleportHQ. This adds to the existing partnership with the Cluj-based Romanian Institute of Science and Technology which is developing advanced artificial intelligence based methods for automatically generating web pages, which may eventually be available through the teleportHQ platform and power its AI.

Paul Brie sees the automation of front-end development as a future mark for the city of Cluj, which could become the main hub of innovation on this topic in the world in the following years: “We could say that in Cluj there’s going to be the biggest number of people in the world that are trying to solve the problem of automated front-end development. So we believe there is a huge chance that Cluj will be known as the capital of automatic front-end development, solved by machine learning. As we make progress in these discussions, more international experts will join the efforts and recognize what we are doing as principal efforts in the industry. So I’m quite confident that in a year or two from now, what we are doing here in Cluj to solve this problem will make history.”

The post Cluj-based teleportHQ launches an ‘AI-wired’ platform for automated user interface generation, setting the stage for Cluj to become one of the main hubs of front-end automation in the world appeared first on Startups&TheCity.