Deep learning audio classifier. 80% F1-score on UrbanSound8K with 10-fold cross-validation. ~20% faster CPU inference via ONNX optimization.

RAG system for financial document Q&A. Hybrid retrieval (semantic + BM25) with ChromaDB, page-level citations. FastAPI + React full-stack.

Deep dive into building a production RAG pipeline — hybrid retrieval, chunking strategies, and page-level citations for financial PDFs.

Tackling extreme class imbalance (0.17% fraud rate) with XGBoost — from feature engineering to deploying a 97% recall model.

End-to-end ML engineering: regression, classification, decision trees, XGBoost, deep learning (TensorFlow/PyTorch), model deployment with Docker, FastAPI, Kubernetes, and AWS Lambda.

EDA, data wrangling and visualization, classical ML (supervised/unsupervised learning), deep learning (neural nets, computer vision), web development, and deployment.

Contributing to financial solutions with a focus on supervised ML techniques. Deploying ML models in production using FastAPI and Docker.

Built end-to-end web scraping pipelines for immigration research. Collaborated on a flood prediction project with NAIRS using time-series modeling for early warning systems.