I build production-grade AI systems — RAG pipelines, multi-agent orchestration, and high-throughput inference infrastructure. Not wrappers. Actual systems.
I'm a full-stack engineer based in Hyderabad, India with 7+ years across healthcare, real estate data, logistics, and enterprise software. I've shipped Redis caching systems that cut query times by 40%, HIPAA-compliant healthcare platforms, and CI/CD pipelines that halved deployment overhead.
Now I'm going deep on AI — building RAG pipelines, LangGraph multi-agent systems, and LLM inference infrastructure. Not as a pivot, but as the next evolution: production engineering experience applied to the hardest problems in AI deployment.
Education: MCA — Netaji Subhash Engineering College · BCA — Global Institute of Science & Technology. Open to relocate. Targeting Senior SDE / AI Engineer roles.
Production FastAPI backend with LangChain document ingestion, FAISS vector indexing, and multi-tenant session isolation. Upload PDFs or text files, query them with natural language, and get answers with full source attribution.
FAANG-portfolio-grade URL shortener with real-time analytics. Base62 encoding, Redis token-bucket rate limiting, serverless AWS Lambda redirects, DynamoDB persistence, and a Next.js dashboard with WebSocket live stats.
Python Playwright bot that automates LinkedIn Easy Apply end-to-end — handles bilingual UI (EN/AR LinkedIn language flip bug), fills dynamic forms, detects success states, and logs every application to CSV. 29/29 unit tests passing.
Upgrading the LangChain RAG project into a full-stack platform: hybrid dense+sparse retrieval with pgvector/Pinecone, HyDE + Cohere reranking, RAGAS evaluation dashboard, and multi-user corpus management.
SaaS platform where users define autonomous AI workflows via a visual builder. LangGraph supervisor + specialist agents execute tasks with persistent checkpointing, streaming token output, and real-time agent status monitoring.
OpenAI-compatible inference gateway with semantic routing (local Ollama vs cloud LLMs), layered caching, vLLM PagedAttention serving, and a React ops dashboard with Prometheus metrics and Kubernetes HPA scaling.
How HyDE + multi-query retrieval + Cohere reranking pushed our RAGAS faithfulness from 0.61 to 0.91 — and what we measured to prove it.
Read →Why checkpointing matters more than the model, and how we achieved 99.2% run success rate on a multi-tenant agentic SaaS platform.
Read →How embedding-based routing cut our cloud inference bill by 82% without degrading response quality. The math, the architecture, and the tradeoffs.
Read →Open to full-time AI/Software Engineering roles, contract projects, and interesting collaborations. I respond to every message — usually within 24 hours.