A day job in enterprise tech pays the bills. This is the part I build because I can't not build it — a health-tech startup, from empty repo to running infrastructure, mistakes included.
Last updated: July 2026 — this is a snapshot, not a live tracker
/internal/photos under the "build" section that matches (01–06) and it'll show up right under that section automatically.
Valensiq AI Healthtech Private Ltd. is a health-tech startup building AI-powered risk stratification for preventive healthcare — the first target is colorectal cancer screening, with a presence spanning India and the UK. I'm CTO, which in a company this size means: I write the code, I design the infrastructure, and when something breaks at 2am, I'm the one who fixes it.
The product is a full-stack build: a Node.js frontend, a Python/Flask backend, and an XGBoost/SHAP machine learning model doing the actual screening prediction — chosen deliberately over a black-box model because SHAP's explainability matters when the output is a health risk score a clinician has to trust.
Running a startup's cloud infrastructure solo — not as part of a platform team, but as the only person who understands the whole system — changes how you build. Every migration, every access policy, every billing account decision is a decision you make alone and live with. Valensiq's GCP setup has been rebuilt more than once as the right architecture became clearer: moving to a dedicated billing account, tightening access down to exactly who needs it, and treating infrastructure as something to actively own rather than something that just runs.
Three things I'm actually solving for right now, not a list of specific services that'll be outdated the moment the architecture shifts:
Underneath all three: keep it cheap enough to run indefinitely and isolated enough — separate environments, separate access, separate everything — to trust with real patient-adjacent data. The specific cloud services behind each of these will keep changing as I learn better ways to do it; the three problems above won't.
Building Valensiq means constantly learning things I didn't already know — cloud platforms, ML deployment patterns, security practices, whatever the current problem demands. One current thread: closing an AWS gap after years of GCP and Azure experience, using the AWS Solutions Architect Associate (SAA-C03) as a structured way to map what I already know onto AWS's equivalents. But that's one thread among several — this section is really about treating learning as an ongoing, visible part of the build, not a one-time credential to collect.
Building isn't only technical. People I meet — on a trip, at work, wherever — sometimes need more than a good conversation, and I've started keeping a proper place for that instead of letting it end when the conversation does.