About
Senior software engineer building at the intersection of large-scale distributed infrastructure and autonomous AI agents. 4+ years at BookMyShow — India's largest entertainment platform — engineering recommendation systems that serve 500M+ users at 500K RPM with 99.99% availability. Currently at Arya Health as a Senior Engineer, building multi-tenant health data pipelines while completing my M.S. CS at NYU (2026). On the side I design agentic systems using MCP, LangChain, and tool-calling LLMs — from AI SRE copilots to autonomous workforce orchestration. Looking to bring both sides — distributed systems at scale and modern agentic AI — to an ambitious team.
Experience
Senior Software Engineer · Data & Platform
- ›Reduced API latency 40% with an Aerospike distributed cache over PostgreSQL
- ›Built serverless event pipelines using AWS Lambda, SQS, and InfluxDB for multi-tenant health data
- ›Wired Dagster pipeline telemetry into SigNoz and Grafana for real-time cross-tenant observability
Software Development Engineer 2 · Personalization & Recommendations
- ›Scaled recommendation systems to 500K RPM at 99.99% availability during blockbuster events
- ›Cut p99 latency 50% (300ms → 150ms) via a Go Scatter-Gather pattern for parallel data aggregation
- ›Led migration from Java monoliths to microservices, reducing AWS infrastructure costs by 40%
Software Development Engineer 1
- ›Built Spark/Scala + Kafka pipelines for zero-loss terabyte-scale cloud migration of user data
- ›Powered real-time recs by integrating Redshift clickstream data into Elasticsearch and Aerospike
- ›Maintained 100% uptime through JVM tuning and Elastic APM alerts during post-pandemic traffic surges
AI / Agentic Projects
Autonomous agents, MCP servers, RAG pipelines, and LLM-powered systems.
Currently Building
Niche AI infrastructure projects — in active development.
Agent combining PostGIS + OpenStreetMap spatial data with LLM reasoning to answer location-aware queries. Explores geo-spatial retrieval as a first-class RAG layer — 'best hospitals near X accepting Y insurance with Z rating'.
API gateway purpose-built for the MCP ecosystem — auth, rate limiting, routing across multiple MCP servers, and built-in observability. Think Nginx/Kong but for agentic tool infrastructure.
Evaluation harness for multi-agent systems. Runs deterministic scenario suites, scores task completion, tracks token cost and latency distribution, and detects behavioral drift across model versions.
AI Focus
The problems I think about most deeply.
Agentic Pipelines
Multi-step LLM workflows — tool use, planning loops, state management, and human-in-the-loop approval gates for destructive actions.
RAG Infrastructure
Vector retrieval, context assembly, hybrid search, chunking strategy, and eval-driven quality loops for grounded generation at scale.
MCP & Tool Ecosystems
Designing, composing, and operating Model Context Protocol servers — auth, routing, observability — for production agentic systems.
Writing
Scatter-Gather at 500K RPM: parallelizing BookMyShow recommendations in Go
How we cut p99 latency in half using a scatter-gather pattern across 12 microservices — and what the Go runtime taught us.
MCP in production: what nobody tells you about building tool servers
Auth, versioning, error taxonomy, and why your MCP server will time out when you least expect it.
Why RAG evals are broken — and a simple harness that actually works
Most RAG eval frameworks measure the wrong thing. Here's what to measure instead, and how to wire it into CI.
Education
M.S. Computer Science · GPA 3.77 / 4.0
B.Tech Computer Science · CGPA 8.06 / 10.0
Stack