Architecting Scalable Distributed Systems | Agentic AI Workflows | AWS Serverless
I am a Senior Application Engineer with 4.5+ years of experience specializing in high-performance backend infrastructure. Currently, I focus on bridging the gap between Reliable Distributed Systems and Probabilistic AI Models.
- Scale: Architected event-driven workflows handling 40K+ daily transactions and 2.5M+ annual claims at Tiger Analytics.
- AI Engineering: Moving beyond simple RAG to build Agentic Workflows (LangGraph) and Self-Correcting Systems that solve actual business problems.
- Location: India 🇮🇳 (Remote Ready & Experienced with Global Teams)
I don't just use tools; I choose the right tool for the job.
| Domain | Technologies |
|---|---|
| Core Backend | |
| Data & Cache |  |
| Infrastructure |  |
| Agentic AI |
A self-correcting knowledge engine for complex decision support.
- Architecture: Hybrid Search (BM25 + Vector) using PostgreSQL pgvector.
- Agentic Workflow: Implemented a LangGraph state machine where a "Critic Agent" evaluates retrieval quality before generating answers.
- Reliability: Integrated RAGAS for automated evaluation of Faithfulness and Answer Relevance in CI/CD.
A unified interface to manage LLM providers with enterprise controls.
- Tech: FastAPI, Redis (Token Bucket Rate Limiting), Celery (Async Logging).
- Features: Implemented Model Context Protocol (MCP) for standardized tool use and fallback logic (OpenAI → Azure → Ollama) to optimize cost vs. latency.
How I approach software design:
- Schema First: APIs are contracts. I define OpenAPIs/Pydantic models before writing a single line of logic.
- Observability is Mandatory: If it's not logged in CloudWatch/Datadog with a correlation ID, it didn't happen.
- Boring Technology: I prefer proven, stable tech (Postgres/Django) for the core, and limit "experimental" tech to the edges.
"Code that works on Friday evening is good. Code that works on Monday morning is better."