The Story of Voxel51: Building the Foundation for AI Development

From academia to AI infrastructure: How Voxel51 evolved from a computer vision research project to building essential developer tools for the future of unstructured AI development.

Written By: supervisor

0

The Story of Voxel51: Building the Foundation for AI Development

The Story of Voxel51: Building the Foundation for AI Development

Sometimes the most impactful companies start not with a grand vision, but with a simple realization about what’s missing. In a recent episode of Category Visionaries, Jason Corso shared how Voxel51 evolved from academic research to essential AI infrastructure.

From Academia to Entrepreneurship

The story begins in a University of Michigan classroom, where Jason, a computer vision professor, met his future co-founder Brian Moore. “He was one of those students who sits in the front of your class and, like, really makes it hard and interesting,” Jason recalls. “Hard in a good way to be a faculty member. Right. Always asking good questions.”

After graduation, they stayed in touch, initially reconnecting over an auto-grader that Moore had written for the computer science department. But their conversations soon turned to bigger possibilities. “I think we were just chatting one afternoon, thinking, like, there’s just so much video out there,” Jason shares. “I think in my office at the time, there were something like six video cameras.”

The First Pivot

Their initial venture wasn’t Voxel51 as we know it today. With a grant from NIST, they started as consultants, building custom video understanding systems. They landed impressive early clients, including the Baltimore City Police Department and companies in the automotive and insurance sectors.

But something wasn’t working. “When it came time to go and sign a long term production, like, let’s get some business intelligence or production value out of this video understanding system, the users just weren’t ready,” Jason explains. For each new client, they “had to spend months getting data, labeling the data, like training models on it, and then were able to get insights. It really didn’t generalize that well.”

Finding Their True Purpose

The breakthrough came when they realized their internal tools were more valuable than their consulting solutions. Instead of building custom systems for each client, they could provide the infrastructure that all AI developers needed.

This wasn’t the easy path. As Jason notes, “It’s not like a sexy thing like data set quality or whatever, but we think of it as critical. Like roadways in American cities are not sexy either, but you need them to get around.”

Building in Public

In August 2020, they made a bold move: releasing their core technology as open source. Initial traction was modest – “In the first months, we had like 20 weekly active users was like a big number for us,” Jason recalls. But they stayed committed to authenticity, refusing to play the hype game common in AI. “We never talk about vaporware. We never promise things in our customer discussions,” Jason emphasizes.

The breakthrough came through strategic integrations, particularly becoming “the preferred mechanism for accessing a dataset called Google open images.” This helped them build credibility and grow their user base organically.

The Future of AI Infrastructure

Looking ahead, Voxel51 sees a future where AI development becomes more dynamic and interconnected. “Data is the heart of machine learning and AI, unstructured data. It’s the essence,” Jason explains. “It’s how we take a capability in the form of an algorithm or a model and we render it down or distill it down to functionality that actually works in a domain.”

Their vision extends beyond just providing tools. They’re building for a world where AI development becomes a continuous feedback loop between engineers and end users. As Jason describes it: “The future of AI that Voxel will support is this world of data apps that are constantly evolving and thriving and connecting, like engineers who are sharing workflows and working on co-development of models and data sets… with their end users through these apps and then getting the feedback over the long haul.”

This future isn’t just about technical capabilities – it’s about enabling better collaboration and understanding between those building AI systems and those using them. It’s a vision where infrastructure isn’t just plumbing, but the foundation for more effective, responsive, and reliable AI development.

Leave a Reply

Your email address will not be published. Required fields are marked *

Write a comment...