A short account.

I am a final-year Computer Science student at CMR College of Engineering & Technology in Hyderabad, graduating in 2026. That is the shortest true sentence about me, and also the least interesting one, so here is the rest.

The summer that changed the shape of things

In 2025 I was selected as a Google Summer of Code contributor with VideoLAN — the organisation behind VLC — to build an AI-powered media recommendation engine. For five months I worked inside a C/C++ and Qt/QML codebase that is older than my degree, alongside a distributed mentor team, in a review culture where nothing lands because you say it works.

That project rearranged how I think about software. Writing a model that scores well in a notebook is one problem. Making it live inside an application that millions of people already have installed — without regressing startup time, without breaking a platform you cannot test on, without asking users to care that it exists — is a different one, and it is the one I got interested in.

What I actually work on

Two threads, and they keep converging. The first is machine learning with consequences attached: a waterborne-disease model trained on 5.25 million calibrated records; a medical vision-language model fine-tuned with LoRA that shows you its own attention map before you trust it; a blood-group classifier built on EfficientNet-B3 with attention, because the cheap version of that test is a real problem worth solving.

The second is the unglamorous work that decides whether any of it survives: the Flask API in front of the model, the PostgreSQL query that was doing a sequential scan, the Docker image, the Nginx config, the CI gate. I am not a researcher who throws artefacts over a wall. Every model in the catalogue is behind an interface somebody can use.

Three hackathons taught me the other half: 2nd Runner-Up at HackByte, an Honorable Mention in the SDG 11 track at IIT Hyderabad’s Hack4SDG among more than two hundred projects, and a finalist place at a 36-hour Specathon. Building under a clock with people you met that morning is an unreasonably good way to learn what parts of your process are real.

What I am looking for

Full-time systems, machine learning, or infrastructure engineering from 2026 — ideally somewhere the model and the thing serving the model are the same person’s problem. I like being close to production. I like being told the measurement is wrong.

If that sounds like something you are hiring for, the form on the front page reaches me directly, or bhargavadepu@outlook.com works just as well.