I'm a freelance AI engineer and ML/DL researcher. My research is on deep learning for time series and sequential data: hybrid forecasting models, and how sequence models behave and can be updated when their data changes over time. I also build AI agents and LLM applications that run in production.
- Hybrid Deep Learning Approach for Load Profile Prediction, STI 2024. BiLSTM/BiGRU/TCN with self-attention and an XGBoost meta-learner for smart-grid load forecasting. Paper · Code
- Enhancing Prediction Accuracy using Ensemble Deep Learning Methods on Time Series Data, ICISET 2024. Paper · Code
- Audit Before You Update: recording artifacts and the limits of bounded model updates in MIMIC-IV ICU prediction. Preprint and code: DataAudit
- Companion study on explanation stability when part of the model is frozen: bounded_update
- Kotha AI: Bangla voice receptionist that books real clinic appointments over a call
- AI Messenger: chat with a family member's AI twin (demo logins in the repo; the free server can take a minute to wake up)
Python, PyTorch, TensorFlow, scikit-learn, XGBoost, OpenCV, Hugging Face, LangChain, LiveKit, FastAPI, Next.js, PostgreSQL


