Bringing Platform Experience to Computer Vision

What do enterprise service management and computer vision have in common?

Well for one, they are now both listed on my resume. I’ve become a very lightly-seasoned computer vision veteran over the past few week. I’m already identifying parallels between my life working with service management orchestration tools and the challenges of building computer vision systems.

Most of the past 14 years of my career involved building or integrating with ServiceNow’s enterprise application platform-as-a-service. I was lucky enough to be part of the company at an early stage and helped grow many teams as we scaled. I was also able to witness the maturity of an entire industry increase with the availability of such a powerful platform.

The part of my job that excited me the most was seeing customers leverage the platform to build applications and automation across multiple systems. The ability to quickly validate proof-of-concept, test, build and iterate until you have a full blown solution makes any developer’s job easier. It also allows you to focus on solving the right problems with the best solution.

Been there…done that.

After 14 years, I’m ready to take on a new market and new challenges.

Machine learning and artificial intelligence have been a passion of mine for a few years. To start my search I talked to early-stage AI companies to find the types of problems that excited me the most.

I talked to a few computer vision related companies solving various problems and they all sounded interesting, and would definitely provide an opportunity for personal growth. Part of my findings from these discussions and the research I did was realizing that building end-to-end CV solutions are complex.

Besides ServiceNow, I also spent some time working with enterprise AI companies. So I’m already familiar with the typical challenges of coordinating between product, data science, engineering, and infrastructure teams. Part of this challenge is coordinating build and release schedules between models and the code implementing the models. Deployment environments are usually predictable and limited. So there usually aren’t any surprises during deployment.

Computer vision use-cases can complicate the deployment predictability. It’s common for vision scientists to spend weeks to months training and optimizing models in a controlled environment. The value of this hard work risks getting lost if the high performing model can’t meet throughput and latency requirements when deployed to target devices. The model is then sent back to the vision team to re-engineer and the cycle repeats.

This is where FØCAL comes in. I recently joined the team at FØCAL because they are defining new development tooling and practices for computer vision. The founder, Brian Rossa, has years of experience as a computer vision consultant building complex systems. His expertise and developer war wounds together with the enterprise-grade platform and orchestration advantages that I’ve built, we’re streamlining computer vision application development.

Product, vision, and engineering teams can now manage the full lifecycle of application development. Annotate, train, test, and deploy computer vision applications from a single platform.

We’re accepting registrations for our early access release coming in September. Let us know what you’re working on and we’ll let you know how we can help. Sign up at f0cal.com.

Not ready to build? I’m happy to chat about CV anytime, just ping me.