Partialfit (Simranjeet.)

I do backend and infrastructure engineering for early-stage startups – Unify, Reddy, Glaze, AdeptMind, Anakin. I like working on hard problems in small teams – the kind where your infra bill matters and your iteration speed matters more.

Cloud costs down 30% at Anakin. GenAI eval pipeline at Reddy that moved QA metrics 20%. Rebuilt Glaze's data ingestion – feed freshness went 4x. These aren't resume lines. You're either already dealing with these problems or you will be.

I'm not for every team. I work best when there's a real problem, a small team that moves fast, and a founder who'd rather ship than slide-deck. If that's you – let's talk.

I've also built things for Caryn AI and TuringMind, contributed to open source, and done research at MIDAS Lab.

What I actually work with: Python, FastAPI, Flask, Django, AWS, Docker, PostgreSQL, Redis, Celery, CI/CD, distributed systems. The kind of stack you use when you're building something real and you can't afford to be clever.

You can reach me at simran@partialfit.com, or on GitHub.

worked with

partial_fit() is a scikit-learn method that trains a model incrementally – in chunks, not all at once. No full retraining. Just continuous, precise improvement.

That's the idea. You don't need a full time ML engineer. You need someone who fits in fast, does the work, and doesn't burn your runway. That's Partialfit.