🌍 I combine AI, physics, and geospatial data to build forecasting and risk models with real-world impact — especially in climate-sensitive sectors like insurance and reinsurance.
- Develop scalable ML pipelines for weather and climate forecasting (daily & sub-daily)
- Design hybrid physical + data-driven loss functions to ensure forecasts respect real-world constraints
- Build models that matter for risk: flood, basin rainfall response, climate extremes
- Work with station data, reanalysis, remote sensing, and hydrometeorological inputs
- Enable reproducibility & explainability: modular code, transparent preprocessing, validation, and deployment
| Repo | Purpose | Highlights |
|---|---|---|
| large-basin-ann-rain-response-pipeline | Annual rainfall response modeling for large basins | Focus on hydrological risk modeling; data-driven + physics-aware approach |
| lstm-weather-pipeline | Station-level weather forecasting using LSTMs | Smart gap-filling, physics-aware loss functions, ensemble & CV support |
| physaware-lstm-forecast | Enhanced version with physical consistency (temperature, dew point, RH etc.) and robust scaling | Designed for applications in insurance / climate risk prediction |
- Languages & Frameworks: Python, PyTorch, Pandas, NumPy, Matplotlib, Seaborn
- Data Processing: Climatology-based gap filling, interpolation, feature scaling, time-series aggregation
- Modeling: LSTM, ensemble methods, loss function engineering
- Domain Knowledge: Meteorology, hydrology, climate extremes, basin rainfall response
- Reproducibility & Collaboration: Modular architecture, clear documentation, validation reports
- Ph.D. Candidate — Centre for Atmospheric & Oceanic Sciences (CAOS), IISc, Bangalore
Research: Decadal atmospheric modes & their influence on local weather extremes - M.Sc. Physics — Banaras Hindu University (2016)
- B.Sc. Physics — University of Calcutta (2013)
📧 Email: [email protected]
- Forecasting climate extremes & basin rainfall with explainable models helps underwriters quantify risk more precisely
- Transparent pipelines enable auditability & regulatory compliance (important for climate risk disclosure)
- Ensemble modeling + physical consistency helps with worst-case scenarios & tail risk — very relevant for catastrophe funds, reinsurance treaties
Thank you for visiting!
⌨️ I believe climate risk models must be as robust, transparent, and physically grounded as they are predictive. If you share that view — let’s connect.