Backend & AI systems engineer building agent runtimes, RAG infrastructure, workflow orchestration, and production backends.

Website | Contact | LinkedIn | GitHub

I like the part of AI products that has to keep working after the demo: durable execution state, scoped tools, retries, memory, tenant isolation, evals, observability, and deployment paths that survive real users.

Right now I work as an Engineering Lead, Backend & AI Infrastructure at Autonomix Solutions, where I build and operate backend-heavy AI systems across agents, retrieval, real-time workflows, infrastructure, and product delivery.

• Agent runtimes and harnesses: tool execution, context management, memory, delegation, approvals, retries, long-running workflows, and human-in-the-loop paths.
• RAG and knowledge systems: multi-tenant ingestion, semantic and keyword retrieval, vector indexes, scoped memory, fault-tolerant embedding jobs, and retrieval quality checks.
• Production backends: APIs, workers, queues, WebSockets, WebRTC, database design, monitoring, rollback, CI/CD, and cloud deployments.
• Product systems end to end: from ambiguous requirements to architecture, implementation, debugging, performance work, and launch.

• Built the core harness for a production multi-agent platform from scratch inside a small engineering team.
• Reworked agent execution around stateless workers, durable run state, idempotent operations, execution journals, and retry-safe recovery.
• Operate 10 concurrent production services across VM fleets, containers, and serverless paths.
• Reduced a critical API path from about 7 seconds to 400 ms through query-plan analysis, indexing, request-path refactoring, and caching.
• Improved a Next.js application by roughly 50% with bundle analysis, dynamic imports, rendering optimization, and caching changes.
• Learned Java/Spring Boot in about three weeks to help lead a four-month enterprise delivery.

• Diet Paglu: AI-powered nutrition tracking for Indian food, built around low-friction meal logging through photos, voice, chat, and WhatsApp.
• ScamGuardian: Chrome extension for spotting potentially fraudulent websites using URL analysis, scanning APIs, and warning flows.
• Modern Developer Toolkit: curated map of modern tools across AI, databases, infrastructure, deployment, observability, and product engineering.

• Languages: TypeScript, Python, Java, JavaScript
• Backend: Bun, Node.js, Hono, ElysiaJS, FastAPI, Express, Spring Boot
• Data and async: PostgreSQL, pgvector, MongoDB, Redis, Weaviate, RabbitMQ, Amazon SQS, background workers, webhooks
• AI systems: agent harnesses, tool orchestration, RAG, embeddings, memory, evals, model routing, token/cost visibility
• Infra: AWS, GCP, Docker, Linux, GitHub Actions, CI/CD, VM fleet management, monitoring, rollback

• Simple systems that are easy to reason about under pressure.
• Explicit boundaries over implicit magic.
• Observability before guesswork.
• Latency, cost, and failure modes as first-class product concerns.
• Shipping the boring reliability work that makes ambitious products feel calm.