I build AI agents that don't just answer — they reason, critique, and act. From multi-stage LangGraph agents that red-team their own decisions, to RAG pipelines with zero-downtime failover, I like shipping systems that are actually live and load-tested, not just notebooks.
- 🔭 Currently building: Multi-agent systems with self-critique loops (LangGraph)
- 🌱 Currently learning: Advanced System Design & Distributed Systems
- 🏆 Top 5% rank, NPTEL Cloud Computing
- 💡 200+ problems solved across LeetCode & GeeksforGeeks
🧠 AI / ML / Agents
LangGraph RAG Pipelines LLM Integration (Groq, Llama, HuggingFace) Embeddings Vector Search (Qdrant) NLP MLOps Statistical Analysis scikit-learn Pandas NumPy
⚙️ Backend & DevOps
FastAPI Spring Boot Docker Kubernetes Jenkins CI/CD GitHub Actions Prometheus Grafana Maven
🗄️ Databases
MySQL MongoDB Qdrant (Vector DB) SQLite PostgreSQL
|
🕵️ The Dossier — AI Investment Research Agent Takes a company name → researches it live → drafts an Invest/Pass verdict → red-teams its own reasoning in an adversarial pass → revises to a final verdict. Includes a Track Record feature that checks past verdicts against real price movement.
|
📊 Xeno — AI Marketing Strategist An AI-native marketing platform — enter a business goal, and an agent analyzes customer data, builds Shopper DNA clusters, recommends & generates campaigns, launches them, and tracks callbacks live.
|
|
📄 Doctel — Multi-Modal AI Document Intelligence Production RAG system with Parent-Document Retrieval for precision search without context truncation. Sub-second semantic search across multi-PDF corpora, deployed across 4 cloud services with 0-downtime inference.
|
💼 Job-Jugad — AI Job Recommendation Platform NLP-powered resume-JD match scoring with missing-skill suggestions. Full MLOps-style CI/CD — Jenkins + GitHub Actions automate build, test, and deploy on every commit, with live Prometheus/Grafana observability.
|
-blue?style=flat-square){kind=icon}
