The original vision for Lightworks was completely different. Inspired by projects like Deepnight, I set out to build an AI model capable of generating MRI-quality imaging from standard ultrasounds. The goal was massive: 10x the savings and accessibility for medical imaging.

But building in healthcare means building for compliance. A model of that nature is regulated by the FDA as an AI Software as a Medical Device (SaMD). Before I could even think about eventual regulatory clearance, I needed a way to build the submission binder. I went shopping for a QMS.

The Compliance Chasm

It became obvious very quickly that the QMS category was entirely legacy, save for maybe Ketryx.

Nothing on the market was built for the people actually writing the software or training AI models. Existing tools felt completely disconnected from modern engineering workflows. They forced developers to leave their environments, ignored local-first workflows, and essentially locked away your data.

Faced with this, I did what any engineer does when the tooling doesn't exist: I built it myself.

The Internal Tool That Became the Product

I built a QMS that I actually wanted to use. The design DNA was simple: 60% Notion, 30% GitHub.

The core philosophy was that design history is version history and a PR is a change request. I extended my GitHub repo and built a system that stored and synced the QMS directly in the local clone. This meant developers could work exactly how they were used to, with zero risk of vendor lock-in because the data was literally theirs, living in their repo.

With the QMS in place, I moved on to sourcing training data for the proof-of-concept ultrasound model. I partnered with imaging centers in the East Bay, ran the math, and hit a wall. Getting the scale of data necessary for the model would require raising $10M+, waiting 18 years for theit organic volume, or something on that spectrum.

That's when the pivot happened. I realized the internal QMS I had built to survive the compliance process and earn regulatory clearance was fundamentally different from everything else in the legacy QMS market.

I built the QMS I wished existed, and realized that it was the company.

The Architecture of Lightworks

Lightworks took compliance, the category everyone avoids, and made it feel like something engineers actually want to open.

Today, Lightworks serves two distinct users:

Quality Managers: They use the Lightworks app directly to maintain rigorous compliance, opening it first thing on Monday morning.

Engineers: They interact with Lightworks less directly, right from their flow state, in their code editors and terminals, through the API, MCP, GitHub, Linear, and Jira integrations.

Big Hairy Audacious Goal

I am incredibly bullish on AI regulation. The FDA's process is the clearest model and precedent we have for safe, effective technology.

Aligning AI development with those rigorous standards is my BHAG. Lightworks is making sure the next generation of builders can achieve that compliance without sacrificing their engineering culture.

Lightworks QMS is Phase 1. Ultrasound to MRI is Phase 2.