A Retrieval-Augmented Generation (RAG) system that grounds LLM responses in verified academic syllabi and engineering textbooks using dense sentence embeddings, FAISS vector search, and high-speed Groq LPU token streaming.
End-to-end applied data science and machine learning system for financial records. Detects spending outliers via unsupervised Isolation Forests, classifies merchant descriptions with supervised NLP models, identifies recurring commitments, and forecasts monthly cash flow.
Full-stack predictive maintenance platform streaming real-time vibration, thermal, and rotational sensor telemetry (ESP32 / simulator) to a FastAPI backend, running Scikit-learn RandomForest anomaly classifiers and a min-heap priority queue scheduler.
Open-source, privacy-first AI browser and local LLM workspace unifying web browsing, on-device model inference (Ollama / llama.cpp), automated DOM context grounding, and client-side AES-GCM encrypted personal vaults (.tea).
AI-driven invoice recovery platform featuring an authoritative FastAPI double-entry accounting ledger, a conversational installment negotiation agent, cryptographic Razorpay webhook verification, and a cross-platform Flutter client.
Messaging automation infrastructure built on Meta’s Graph APIs, featuring real-time webhook ingestion, keyword intent matching, comment-to-DM pipelines, and containerized microservices behind Traefik.
Structured educational resource hub and syllabus indexing platform built with Next.js, TypeScript, and Supabase for fast curriculum navigation and revision tracking.
