My Projects
A selection of projects demonstrating my skills in data science, machine learning, and cloud technologies. Hover for a preview (desktop). Click to learn more.
LLM Concepts RAG System
Advanced Retrieval-Augmented Generation for Machine Learning Education
An end-to-end RAG application that answers ML questions using hybrid search (BM25 + Vector) and cross-encoder reranking. It features dual-LLM fallback (Groq & OpenAI), dynamic Mermaid diagram generation for visual learning, and a fully Dockerized architecture with PostgreSQL & Grafana for monitoring interactions.
SaarthiEO Disaster Detection
AI-powered disaster classification from satellite & drone imagery
An end-to-end AI pipeline for classifying natural and man-made disasters (Fire, Flood, Collapsed Building, Traffic Incident). Features Grad-CAM explainability, real-time inference, and a production-grade web interface. Built for rapid first-responder triage.
AetherRead
AI-powered, offline-first PDF reader & research workspace
A production-grade, offline-first PDF reader and AI research workspace. Lets students, researchers, and academics read, annotate, and interact with documents entirely from their local machine — no internet, no cloud dependency, zero data sent anywhere. Features 6 reading themes, page-level annotations, SHA-256 duplicate detection, and reading state persistence.
British Airways Predictive Analytics
Tackled real-world challenges impacting airport operations and customer acquisition strategy. Engineered a reusable lookup table to forecast lounge demand, and built an ML classification model (XGBoost) to predict customer buying behavior with actionable feature importances.
Python · Pandas · Scikit-Learn · XGBoost
CBA InsightSpark Data Platform
Built a robust data platform capable of ingesting, processing, anonymizing, and modeling data from diverse sources. Aggregated transactional datasets, enforced strict data privacy (PII masking & hashing), and designed normalized relational databases for Twitter data using SQLAlchemy ORM.
Python · SQL · SQLAlchemy · PII Anonymization
Anuj's Notes
Interactive German A1 learning notes & flashcard system with spaced repetition
A personal language learning platform with detailed German A1 episode notes, custom 3D flip flashcards, built-in spaced repetition tracking, and a fully responsive design — built from scratch with Vanilla JS and CSS.