Back to Podcast
Podcast February 12, 2026

The OneTrack Story: From Drones to AI-Powered Warehouse Intelligence

Featuring Marc Gyöngyösi, Founder and CEO of OneTrack

Marc Gyöngyösi joins Noah Labhart on the Code Story podcast to tell the full creation story of OneTrack — from building autonomous drones in a freezing Chicago warehouse and crashing one on stage at TechCrunch Disrupt, to building an AI agent platform that helps some of the world's largest corporations run their operations.

About the Episode

This is a special guest appearance on Code Story, a podcast that interviews tech visionaries about what it really takes to build something from nothing. Marc shares the unfiltered journey behind OneTrack, including the pivots, the hardware mistakes, the early MVPs, and why the company never took venture capital.

Key Topics Discussed

The Origin: Drones in Warehouses

Marc started the company in 2017 as Intelligent Flying Machines, building autonomous drones for warehouse inventory checks. The idea came from his time at BMW in Munich, where misplaced parts in the warehouse caused robots on the production line to be idle, and people were literally running around with binoculars trying to find pallets in rack.

TechCrunch Disrupt and 12 Stitches

The drone crashed on stage at TechCrunch Disrupt in front of everyone. And they didn't know Marc was still in college. It was a pivotal moment, but the real wake-up call came when a drone run left him with 12 stitches. "Turns out it's a really bad idea to put spinning blades next to people."

The Pivot to Computer Vision

Instead of flying robots, Marc applied the same core perception technology to a broader computer vision system. Cameras on forklifts capture what actually happens on the warehouse floor, capturing signals about people, process, product, and customers that legacy systems could never see.

The Age of MVPs

OneTrack doesn't believe in one big MVP. They stack small wins and move fast. From the first drone demo in a freezing Chicago warehouse, to the first full site deployment with 34 sensors, to a safety event chat box that evolved into a full AI agent platform, every day is an opportunity for a new MVP.

Building Agentic AI for the Physical World

The last 12 months have been a breakthrough. OneTrack captures the physical-world data that agentic AI needs to actually automate warehouse operations. You can't automate a warehouse running on green screen systems with partially accurate data. OneTrack brings in the truth of what's actually happening, then layers on an agent platform that can do in seconds what used to take hours, days, or weeks.

Scalability at Billions of Images

OneTrack processes billions of images every single day, condensing them down to thousands of events, then a handful of insights that change how customers operate. Training self-supervised vision models at this scale requires careful architecture decisions about storage, indexing, and access.

Agent-First Application Design

Marc describes a fundamental flip in thinking: instead of building an application by humans and then trying to automate it with agents, OneTrack builds applications by agents that are used by humans. The same API layer serves users, services, and agents equally, so expanding into new domains like quality, safety, or optimization doesn't require backend rework.

Building a Business, Not a Startup

OneTrack has never taken venture capital. Marc credits mentors like Gordon Segal, the founder of Crate and Barrel, with shaping his long-term perspective. "We're building a business," not riding the VC hamster wheel of raising every 12-18 months. That approach creates a different kind of relationship with customers: partners, not vendors.

The Future: Enterprise Software That Understands the Physical World

Marc envisions a future where enterprise software natively understands the physical world. No more paper scanning, form filling, or digging through 40-message email threads. The enterprise software stack of the future will combine physical-world intelligence with AI agents that don't just automate back-office tasks but give people the dynamic, adaptive software they actually need.

Key Quotes

"We live in the age of MVPs. There's not just one MVP. Stack a lot of small wins and move really fast."

"Turns out it's a really bad idea to put spinning blades next to people."

"You can't really automate warehouse operations if you're running on data in a green screen system that is only partially accurate."

"We flipped the concept around: from trying to automate an application built by humans with agents, to having an application built by agents used by humans."

"The most important thing is everyone here is driven by the impact they have on the customer."

"Really understand the business you're building and the why of why your customers buy what you're making, and be honest about that with yourself."

"I have a very long-term view on what I want to build with OneTrack. I want to be the first person they call when they're going down that next journey."

See the vision Marc describes in action.
AiOn is OneTrack's AI agent platform, built to connect the physical world to digital intelligence and automate operations from the ground up.

Explore AiOn →

Key Takeaways

  1. Physical-world data is the foundation - You can't build AI for operations on inaccurate legacy data. Start with what's actually happening.
  2. Agent-first, not agent-added - Build applications around AI agents from the start, rather than bolting agents onto human-built software.
  3. Speed compounds - Daily MVPs and rapid iteration beat 12-month roadmaps when the underlying technology moves faster than you can plan for.
  4. Bootstrap with purpose - Not taking VC funding creates a long-term mindset that leads to deeper customer partnerships.
  5. The wall might not be technical - Sometimes the hardest problem isn't the algorithm. It's being honest about whether the problem you're solving is big enough.

Related

Want to see OneTrack in action?

Discover how AI-powered visibility can transform your warehouse operations.

Book a Demo