Startup Challenge 2025: Top 3 Finalists Revealed

Drumroll please: the big moment is here. We’re excited to announce that DeepTempo, Lumilinks and Winning Variant will compete for the top prize at this year’s Dev Day!

These three startups will make their pitch in the Snowflake Startup Challenge finale on June 5, vying for the title of Startup Challenge Winner and a share of up to $1 million in investment from Snowflake Ventures, plus exclusive mentorship and visibility opportunities from NYSE. 

Many thanks to the rest of our semi-finalists for their presentations during the second round of competition. We appreciate your effort and dedication.

Read on to learn more about the 2025 finalists and their solutions, and be sure to register for Dev Day to attend the live finale in San Francisco!

DeepTempo

As data and system security becomes more sophisticated, so do the attempts to break through it. Professional attackers are launching AI-aided attacks that are sneaking past traditional rules- and ML-based cybersecurity tools, even as the costs of operating and maintaining those tools continue to rise. 

DeepTempo picked up the challenge and created Tempo, a Snowflake Native App that uses deep learning-based detection to help customers identify security events and analyze both scope and severity with extreme accuracy.

“We saw that foundation models were great at catching patterns in fraud, so we built a LogLM for cyber threats,” said Evan Powell, Founder and CEO of DeepTempo. “With BNY’s partnership, we trained it on massive real-world data, creating a system that sees attacks others miss — and adapts fast as attackers change.”

DeepTempo’s purpose-built LogLMs, or Log Language Models, are trained on enormous quantities of security log data with a focus on event and behavior patterns. With Snowflake, the DeepTempo team realized that they could take advantage of running on a data lake and the LogLMs’ ability to adapt quickly to directly address the time-to-value challenge that impedes security innovation. The team put its technology to task when it completed the BNY Ascent Program, working with BNY on evaluating DeepTempo through a proof-of-concept (POC) deployment for cybersecurity incident identification.

As an early-stage company, DeepTempo is looking forward to the chance to pitch at Dev Day 2025 — and the opportunity to receive mentorship from an NYSE-listed company. 

“We are very interested in mentorship that will help us to prioritize our development and adjust our marketing as well,” said Powell. “For example, should we sell like most security companies —  top-down through the CISO — or should we partner with the AI and platform teams within the enterprise to offer this sort of security as a service?”

Lumilinks

Imagine you have a big toy box, but it's so messy you can't find your favorite toys. You need help to tidy up the toy box so you can find what you need and have more fun playing. Lumilinks aims to be that helper by building a library of AI-ready applications that enable business teams to unlock their data and use it to solve real industry challenges such as predicting when a fleet vehicle will break down, automating the next best action for a customer, or maximizing the lifetime value of an asset.

“We recognized the opportunity when we saw companies investing heavily into Snowflake, Document AI and Cortex, but struggling to achieve the full value because users lacked the skills, time or tools to engage deeply,” said Gary Cole, CEO of Lumilinks. “By wrapping Snowflake's powerful capabilities into business-ready applications, we help drive real consumption, adoption and business success.”

Lumilinks’ FleetSense AI product targets an area ripe for AI assistance: fleet management. Fleet managers, vehicle operators, procurement and finance each needing different outcomes from the same data, which traditional fleet management systems are hard-pressed to deliver. By bringing predictive intelligence to fleet operations, Lumilinks helps teams identify potential maintenance issues that could cause vehicle downtime, process inefficiencies and stay current with constantly changing regulations. 

The impact of applying AI can be significant. A leading water utility used Lumilinks to help reduce average daily vehicle off-road time by up to 45% with predictive maintenance and unlock a potential £2.6 million in annual value while reducing incident risks and operational disruptions. By leveraging FleetSense AI, they minimized maintenance costs, ensured faster response to critical incidents, and kept their operations flowing smoothly, driving significant ROI and operational resilience. 

The Lumilinks team is excited by the opportunities offered by the Snowflake Startup Challenge and the chance to meet the challenge judges — and possibly pick the brain of Benoit Dageville, Snowflake’s co-founder and President of Product, about his thoughts on commercially-focused AI apps as a driver of AI adoption and the future of Snowflake Marketplace. 

Winning Variant

Testing two ideas side-by-side to see which is better is something all of us have tried — but for developers and product managers, it’s a foundational part of building a better product. Winning Variant aims to provide Snowflake customers with a split testing platform that looks and feels like it was built and customized in-house. That includes complete data security, analytics flexibility and visualization in their existing BI tool.

The drive to make this vision a reality stems from Winning Variant Founder Kirk Morales’ experience after his last startup was acquired. He went on to lead the team that managed the in-house experimentation platform for a Fortune 300 company (and a Snowflake customer), and learned that many other growing companies eventually outgrew their testing solutions or had to deal with a product that couldn’t address the nuances of their business.

“If a company is looking to release a new Cortex Agent or ML model version, Winning Variant would let them test both versions side-by-side to measure the impact to core business metrics,” said Morales. “This allows teams to ensure that new versions will move the business forward — increase revenue, reduce churn, prevent fraud, and so on — before pushing a widespread deployment.”

With Winning Variant’s solution, this same principle of split testing can be applied to websites, apps, SaaS products and even Snowflake workload optimization. Available as a Snowflake Native App, Winning Variant’s experimentation platform allows customers to run innovative experiments directly inside the AI Data Cloud. Teams can tap into data available in Snowflake without having to copy data, access a third-party platform or build additional data pipelines. The result: an agile, efficient way to test new features and facilitate experimentation.

Like the other Snowflake Startup Challenge finalists, Winning Variant is intrigued by the chance to meet other market-leading organizations through the NYSE mentorship opportunity.

“It’s rare to have the opportunity to be mentored by top companies, so it would be a career highlight to have that chance,” said Morales. “Our team believes in constant learning and improvement — who better to provide personal and company growth opportunities than someone at the top of their game?”

Stay tuned for the Startup Challenge Finale in San Francisco

Want to see which of these three startups will take the top prize? Register for Dev Day now to see the live finale and experience all of the developer-centric demos and sessions, discussions, expert Q&As and hands-on labs designed to set you up for AI/ML and app dev success.

It’s also never too soon to start thinking about the next Snowflake Startup Challenge: Complete this form to get notified when the 2026 contest opens later this year.