I build systems that turn raw signal into decisions — anomaly detection, forecasting, and applied NLP, shipped as real production pipelines, not notebooks.
I'm a data scientist based in Maharashtra, India, working at the intersection of machine learning and production engineering. My focus is on systems that need to make correct decisions in real time — fraud and anomaly detection, demand forecasting, and signal extraction from noisy, high-volume data.
Most of my work starts as a research question and ends as a deployed API someone else's product depends on. I care as much about model drift monitoring and latency budgets as I do about validation metrics — a model that's 2% more accurate but can't run in production isn't a win.
Outside of client and product work, I write about applied ML tooling and contribute to a couple of open-source data pipeline libraries.
Real-time anomaly detection platform for network traffic. Streams live events through Kafka, scores them with an ensemble of Isolation Forest and LSTM autoencoders, and flags outliers to a live dashboard before they become incidents. Cut mean-time-to-detection from ~40 min of manual log review to under 90 seconds.
Customer churn prediction pipeline for a subscription SaaS product. XGBoost model with SHAP-based explainability surfaced to account managers, so retention outreach is prioritized by actual risk drivers instead of gut feeling. Deployed as a scheduled batch job on AWS Lambda.
On-device computer vision model that identifies crop disease from a leaf photo, built for low-connectivity rural use. CNN trained on an augmented open-source leaf dataset, quantized and shipped via TensorFlow Lite into a Flutter app usable fully offline.
NLP pipeline that ingests financial news and social sentiment, embeds it with transformer models, and aggregates a daily sentiment signal per ticker. Backtested as a secondary input alongside price-based indicators.
Currently based in Maharashtra, India — open to remote and hybrid opportunities. Reach out directly, or find me on GitHub / LinkedIn.