Machine learning
Forecasting · Classification · Deep learning · NLP · Experiment design
Data Scientist & ML Engineer.
I turn messy data into models that ship.
- 12
- models in production
- 38%
- lower latency
- Python · PyTorch · SQL · MLOps
Selected work
Models that
moved the numbers.
Selected data science and machine learning projects, from problem to production.
A little about me
I care about models that work in production, not only in notebooks.
I am a data scientist and ML engineer who owns the full path from problem framing and experiments to deployment and monitoring. I like simple baselines, honest evaluation, and systems other people can maintain.
✦Currently exploring LLM-powered tools and reliable ML evaluation.
Capabilities
From the first question
to the production model.
Data & engineering
Python · SQL · Spark · Airflow · FastAPI
MLOps & cloud
Docker · Kubernetes · MLflow · CI/CD · AWS
Experience & education
A few places that
shaped my perspective.
2023 — Present
Data Scientist & ML Engineer
Current company
Building and deploying machine learning models, from feature pipelines to production APIs and monitoring.
2021 — 2023
Data Scientist
Previous company
Delivered forecasting and classification models and the analytics that helped teams act on them.
Education
B.E. / B.Tech, Computer Science
Your university · 2017 — 2021
Off the clock
Things that keep my perspective fresh.
Open source
Reading and contributing to the tools I use.
Research papers
One new paper a week, at least.
Long runs
The best kind of mental reset.
Good films
Always building a watchlist.
Writing
I write about data science, ML in production, and lessons from shipping models.
Latest posts
All posts ›Baselines before deep learning
Why the first model I build is almost always the simplest one, and how it keeps every later improvement honest.
Monitoring models in production
A model is only finished when you can tell it is still working. The few checks I set up before every launch.
Evaluating forecasts honestly
A single train/test split flatters time-series models. Rolling backtests give a number you can trust.