Som is having 17+ years of engineering · LangGraph & LangChain Architect · Designing production-grade autonomous systems that reason, route, and retrieve at scale.
Production systems that move beyond demos — built for scale, reliability, and real-world deployment.
A production-grade multi-node LangGraph graph with intelligent query routing between PDF vector retrieval and live DuckDuckGo web crawl. FastAPI backend, SHA-256 deduplication, AstraDB vector storage, and AWS ECS Fargate deployment pipeline.
A Retrieval-Augmented Generation pipeline with conditional edge routing and state machine logic — deciding dynamically between dense vector search, sparse keyword retrieval, and live web augmentation based on query confidence scoring.
End-to-end model fine-tuning using PEFT techniques (LoRA & QLoRA) targeting domain-specific tasks in Banking and E-Commerce. Efficient adapter training with quantization to run on constrained GPU infrastructure without quality loss.
Kong / nginx API gateway integration for multi-agent FastAPI services — rate limiting, JWT validation, circuit breaking, and distributed tracing via OpenTelemetry. Designed to manage agent-to-agent communication in high-throughput environments.
DataStax AstraDB-backed persistent knowledge store with configurable vector dimensions, keyspace partitioning, and real-time similarity search. Integrated with LangGraph state nodes for low-latency document retrieval across enterprise knowledge corpora.
GitLab CI/CD pipeline for containerised LangGraph FastAPI services — automated Docker builds, ECR pushes, ECS task definition updates, and rollback strategies. Includes environment-specific staging gates and integration test hooks.
Every tool chosen for production readiness — not demo impressiveness.
Open to senior Agentic AI Architect opportunities at MNCs and high-growth startups.
Let's discuss what intelligent systems can do for your organisation.