╔══════════════════════════════════════════════════════════════════════════════╗
║ Building production AI systems that move beyond prototypes — ║
║ Graph-RAG engines, agentic pipelines, and full-stack AI platforms ║
║ with measurable reliability, end-to-end observability, and real impact. ║
╚══════════════════════════════════════════════════════════════════════════════╝
I'm Anupam Kumar, an AI Engineer with a B.Tech in Computer Science from IIITM Gwalior (2021–2025). I build production AI systems — Graph-RAG engines, agentic pipelines, and full-stack platforms — designed to hold up under real traffic, not just in a demo.
My work spans three areas I like to own end-to-end: AI systems (retrieval, reasoning, agents), backend & platform engineering (APIs, distributed services, containerized deployment), and ML infrastructure (evaluation, experiment tracking, observability).
Currently open to full-time roles in AI Engineering · Platform Engineering · ML Infrastructure.
Languages
AI / LLM
Retrieval & Vector Search
Backend & Data
DevOps, Cloud & Observability
Frontend
Citation-graph-aware RAG system answering computer vision research questions across 238 papers with zero hallucinations.
Python FastAPI FAISS BM25 Redis PostgreSQL Prometheus MLflow Docker
- Built a production Graph-RAG platform combining dense, lexical, and citation-graph retrieval to reach 0.94 context recall and a 0% hallucination rate across 238 papers / 12,288 chunks.
- Engineered a 7-layer Redis caching system (exact, semantic, retrieval, decomposition, intent) that drove an 84%+ cache hit rate while keeping p99 latency near 4.9s on CPU-only infrastructure.
- Designed a multi-signal semantic router (domain centroid, retrieval-support probe, entity shape) for out-of-domain detection, keeping responses 100% grounded with zero false-positive escapes.
- Instrumented full MLOps observability — MLflow tracking, Prometheus/Grafana monitoring, containerized deployment — with zero evaluation failures across all benchmark runs.
Multi-tenant platform that discovers, enriches, and scores sales leads at scale.
LangChain Playwright Celery FastAPI PostgreSQL React Docker
- Architected a multi-tenant lead-intelligence SaaS processing 1,000+ leads/day via distributed Playwright scrapers running on Celery workers.
- Built a LangChain + Groq enrichment pipeline for semantic entity extraction and contact qualification, cutting manual research effort by up to 80%.
- Implemented a JWT-secured, role-based multi-tenant FastAPI backend with a React dashboard for real-time lead scoring and pipeline analytics.
- Containerized the full stack with Docker Compose for one-command deployment.
Citation-grounded RAG platform making 10,000+ pages of Sanskrit manuscripts searchable and conversational.
FastAPI PyTorch Pinecone LangChain Docker
- Built a custom Char-CNN OCR model to digitize Sanskrit-script manuscripts — solving a problem standard OCR tools fail at entirely.
- Indexed 10,000+ pages into Pinecone, enabling sub-second semantic retrieval across the classical text corpus.
- Reranked candidates with a cross-encoder and generated citation-grounded answers traceable to a specific source passage.
- Powered multi-turn consultations through a LLaMA-3 conversational layer with persistent session memory.
End-to-end automation from LinkedIn job discovery to Easy Apply submission, gated by AI resume scoring.
Python Playwright Streamlit SQLite Groq LangChain
- Automated the full pipeline: LinkedIn scraping → LLM resume scoring (70% semantic, 30% keyword) → Playwright-based Easy Apply submission.
- Enforced a strict state machine (
Discovered → Scored → Queued → Applied/Skipped/Failed) with every decision logged to SQLite. - Built in safety controls — daily application caps, a dry-run mode, and a 4-tier LLM fallback chain (Groq → OpenRouter → HuggingFace → keyword) — for reliable, unattended operation.
- Shipped a real-time Streamlit dashboard for match-score distribution, application trends, and CSV export.
Indian Institute of Information Technology and Management, Gwalior B.Tech in Computer Science | 2021 – 2025
Open to collaborating on AI engineering projects, or discussing full-time opportunities in AI Engineering, Platform Engineering, or ML Infrastructure.