Faizaan Khan

A production-style ML forecasting pipeline for retail demand planning using Apache Airflow, StatsForecast, FastAPI, Streamlit, and Docker Compose. The platform handles preprocessing, scheduled model training, holdout evaluation, artifact promotion, API-based prediction, dashboard monitoring, and CI checks for reliable ML workflow delivery.

A document intelligence backend for multi-format file ingestion, chunking, vector indexing, hybrid vector/BM25 retrieval, and grounded answer generation with source chunks. Includes API-key protection, clean service-layer architecture, Docker support, tests, and production notes for scaling with managed vector databases, API gateways, PostgreSQL metadata storage, logging, and tracing.

A Python-based quantitative research framework combining vectorized Polars preprocessing with an event-driven execution engine for realistic strategy backtesting. Supports technical indicators, multi-asset portfolios, margin, commissions, short selling, position sizing, walk-forward analysis, Monte Carlo testing, optimization workflows, and AMM-aware DeFi execution.

An advanced Retrieval-Augmented Generation system that goes beyond basic chatbot retrieval through a LangGraph-powered agentic workflow. Combines FAISS dense search, BM25 sparse retrieval, reciprocal rank fusion, cross-encoder reranking, corrective query rewriting, optional web-search fallback, multi-turn memory, FastAPI serving, Streamlit UI, Docker deployment, and RAGAs-based evaluation.