Introducing the 2026 Snowflake Startup Challenge Finalists: Airrived, LGND AI and Twine Security

There's a common theme across the 2026 Startup Challenge Finalists: intelligent connection. It takes different forms — an Agentic OS orchestration layer, coordinated agents that tie together data and action, new forms of models that make Earth imagery queryable — but it's clear that AI's potential to reinvent how people and systems interact with data (and each other) is inspiring our three finalists to think big.

We are pleased to announce that Airrived, LGND AI and Twine Security will advance to this year's Snowflake Startup Challenge finale and face off for the opportunity to receive a share of up to $1 million in investment from Snowflake Ventures, plus exclusive mentorship from NYSE-listed companies and the opportunity to ring the NYSE bell.

Many thanks to the other semifinalists for their dedication and the effort they put into their presentations during the previous round.

Read on to learn more about the three startups vying for the top spot in the 2026 competition.

Airrived

Despite large investments in data, tools and infrastructure, many enterprises still rely on manual workflows and piecemeal solutions to get work done. That's the gap Airrived was built to close.

"We saw that the real opportunity wasn't another tool but a foundational layer: an Agentic OS that brings intelligence to data, orchestrates actions across systems, and enables teams to move from insight to execution in real time," says Anurag Gurtu, Co-Founder and CEO of Airrived.

Airrived's agents connect to existing systems rather than replacing them, giving enterprises an intelligent execution layer on top of the infrastructure they've already built. It can help operationalize workflows where an organization needs to move from data to action such as security, operations, compliance and IT management.

For example, employee access governance — managing which employees can access which systems — is a compliance-critical function that usually involves periodic manual reviews. The time it takes managers to comb through spreadsheets slows down access approvals, and risks may go unnoticed.

With Airrived, that process becomes persistent and intelligent. Arrived's agents can run in the background, analyzing access patterns, identifying excessive or risky permissions, and recommending or executing changes with the appropriate approvals. There's no need to wait for the next scheduled review cycle because the process can now run continuously in the background, and governance teams can spend less time on manual reviews and more time on strategic work.

Airrived's team is closely watching how enterprises are moving from AI experimentation to full operationalization, and noted that one of the Snowflake Startup Challenge prizes would be especially useful as they navigate the fast-changing AI world.

"Mentorship from NYSE-listed companies would be invaluable as we scale," says Gurtu. "We're building foundational infrastructure for the agentic era. Learning from companies that have navigated global scale, governance and public market expectations would help us accelerate responsibly while maintaining trust and reliability."

LGND AI

Today's LLMs are trained on language, which means the physical world — and the 800-plus petabytes of Earth imagery that captures it — is largely off-limits to them. LGND AI wants to change that by using AI to help understand the Earth, and make the Earth understandable to AI.

The company builds what it calls Large Earth Observation Models (LEOMs) by applying the same transformer architecture that powers LLMs to pictures of the Earth. The result is a platform that makes the physical world queryable, for both human analysts and AI systems.

The immediate beneficiaries are geospatial analysts and domain experts working in industries such as insurance, agriculture, mining and intelligence. Traditionally, extracting structured insights from satellite imagery meant purchasing images, hiring specialists and spending months training custom computer vision models to get a usable data set. With LGND, a nontechnical analyst can ask a question like "Where is there new industrial construction in Texas this year?" or "How much deforestation happened in the Amazon in the past three months?" and get a structured answer in seconds.

The use cases are broad and varied. An insurance company could run rapid damage assessments across a hurricane-affected region before field surveyors arrive on the ground. An underwriting team could build comprehensive data sets of firebreaks and flood barriers for its risk models in hours instead of months.

The bigger opportunity LGND is pursuing, however, is becoming the geospatial layer for AI. Being able to query the Earth is now an API call away, which opens up a wide range of possibilities, according to LGND's founding team.

"Imagine going to book a hotel and being able to ask 'Is there construction happening near this property?' " says Nathaniel Manning, Co-Founder and CEO of LGND. "We suspect that very soon the largest users of our API won't be developers, but agents querying the Earth to solve a larger problem for a business or an individual."

Twine Security

The inspiration for Twine Security emerged during Nadav Erez's time at Claroty.

"For years, I watched customers deal with the same problem," says Erez, Co-Founder and CTO of Twine Security. "They had no shortage of tools, policies or well-defined programs, but never had enough resources or time to execute them fully."

Twine's answer to this execution shortfall are AI Digital Employees that work alongside cybersecurity team members to carry out projects and tasks that would otherwise pile up in the backlog.

Built using Snowflake Cortex AI, the company's first AI Digital Employee is Alex, who specializes in identity and access management (IAM) — one of the most labor-intensive areas in cybersecurity. Alex learns each organization's environment, surfaces the most pressing security gaps and then takes action to close them by taking care of entire workflows, including solving IAM tickets, cleaning up stale accounts and accelerating user access reviews. A human manager is always in the loop for governance, but Alex is the one doing the actual task execution.

IAM is just the first area Twine is tackling. The company has plans for a broad collection of AI Digital Employees, each trained to take on a different domain of security work. Snowflake's unified data platform brings it all together, giving Twine access to the connected data sources, governance and scalability it needs to bring Alex and other AI Digital Employees to life.

"Agentic AI is only as good as the data it uses," says Erez. "That's why we're building on Snowflake, and why we're excited to go onstage at the Startup Challenge Finale and show what we've created."

Register for Snowflake Summit and Dev Day to see the Startup Challenge finale live and explore over 500 demos, expert sessions, Q&As and hands-on labs designed to jump-start innovative ideas and set you up for success in the era of agents and enterprise intelligence.