AI Engineer · Automation Systems
"Making things that make machines think."
A model that hits 92% in a notebook and breaks in production isn't a win — it's a warning sign.
Every system I build goes all the way: pipeline, training, evaluation, deployment, real users.
If it doesn't ship, it doesn't count.
Things that handle the real world — class imbalance, multimodal signals, and live production traffic.
End-to-end data pipelines built for real-world noise — class imbalance, missing data, adversarial edge cases. Every model ships with explainability baked in.
Audio. Vision. Text. Combining signals to make machines that understand context — not just classify inputs.
FastAPI backends. React frontboards. MongoDB. Docker. I deploy my own models because I want to understand every layer they run on.
I didn't get into AI because it was trending. I got into it because I kept running into problems that existing tools were too dumb, too expensive, or too inflexible to solve. So I built the tools myself.
"Calorie apps made me enter every ingredient by hand."
Photograph food. Get macros instantly. Gemini Vision API. 92% accuracy. 80% less manual work.
"Studying without structure was pure chaos."
Pomodoro tracker with streaks and visual progress. Built it in a weekend. Still use it every day.
"Wanted a quiet way to know if someone around me was struggling."
Speak into it. It reads your emotional state. CNN on audio features. 87% accuracy. Live on Hugging Face.
"Fraud detection systems shouldn't be a black box."
284,807 real transactions. Random Forest + SHAP. Every decision explained. Live on Hugging Face.
Things I've engineered and shipped.
A model that needs a manual isn't smart — it's just complicated. The real work is making something powerful enough to handle the hard parts, and simple enough that you trust it without thinking about it.
That's where I build. Not at the boundary of capability. At the boundary of capability and clarity.
The Manchester of the East — industrial by heritage, Biryani and Thandai by soul, and the Ganga flowing quietly through all of it.
It's a city that doesn't perform. It just works. Directly. Without pretense. That's shaped how I build too: no unnecessary weight, no overengineering, just clear thinking applied to real problems.
Open to AI/ML engineering roles — internships, full-time, freelance. Remote or on-site. Pan-India.
If you're working on something that needs a machine to think, I'd like to hear about it. Over email, or over chai.