Moleculeo
An AI-powered drug discovery and drug-repurposing platform. As Co-founder and CTO, I built the platform from the ground up and own all technology and technical strategy.
Visit moleculeo.comKai McKinnon · New York, NY
I build software products, end to end.
I'm the Co-founder and CTO of Moleculeo, where I built an AI drug-repurposing platform from the ground up. Before that, I spent a decade building clinical research software at Flatiron Health and Medidata — data pipelines, electronic data capture, and the platforms trials run on. Alongside Moleculeo, I take on select one-off projects: new products, data platforms, and stalled builds that need to ship.
01 — Selected projects
An AI-powered drug discovery and drug-repurposing platform. As Co-founder and CTO, I built the platform from the ground up and own all technology and technical strategy.
Visit moleculeo.comA full-stack tutoring platform for a New York City math tutoring nonprofit, used by dozens of weekly tutors, school coordinators, and administrators. I lead its volunteer engineering team.
Visit cdmath.orgThe clinical data platform at Flatiron Health underpinning its clinical research revenue plan — a plan to grow from $12 million to $100 million. I was its sole accountable engineering owner.
Read the case studyElectronic data capture for Roche's Lighthouse study — scaled to a study of more than 100,000 patients, with an end-to-end test suite cut from 16 hours to 2.5.
Read the case study02 — Engagements
Idea to working product: architecture, build, and launch, owned by one person who has done it before. Proof: Moleculeo and cdmath.org, both built from the ground up.
Pipelines, integrations, and data products for research — EHR-to-EDC, study data flows, and the tooling around them. Proof: Clinical Pipe at Flatiron Health.
Applied AI inside real products, not demos — search, matching, and workflow automation that ships to users. Proof: Moleculeo's mechanism-based drug-repurposing search.
When a build is slow or stuck, I find the bottleneck and fix it. Proof: a test suite cut from 16 hours to 2.5, and average cycle time from 33 days to 26.
03 — Process
A short, written scope: what gets built, what done looks like, and a fixed timeline. No open-ended engagements.
Working software every week, in a staging environment you can actually click on — not status reports.
Production launch, documentation, and a clean handover. Your team can run it without me.
Tell me what you're trying to ship. The fastest way to reach me is LinkedIn — or browse the code on GitHub first.