I build production LLM systems: voice agents, multi-agent orchestration, retrieval, and the infrastructure that keeps them running.

• 01

  Harbor · Kubernetes MCP gateway

  LiveCode

  One /mcp endpoint in front of many MCP servers, with per-user keys, audit logging, a chat agent and a dashboard. Helm charts for kind and k3s, plus a custom WooCommerce MCP server.

  • TypeScript
  • Kubernetes
  • Helm
  • Postgres
  • 100+ tests
• 02

  Hotseat · AI sparring partner for high-stakes meetings

  LiveCode

  Builds a persona from the counterpart's profile, predicts objections, runs a streaming role-play with a live stance meter, then scores you with quotes from the conversation.

  • Next.js
  • SSE
  • Multi-provider LLM
  • Turso
  • 90+ tests
• 03

  Dory · WhatsApp health assistant

  Code

  PCOS-focused WhatsApp bot: consent, conversational survey, rule-engine matching, evidence-grounded content, lab-report OCR and pgvector RAG Q&A. Verified end to end on the Meta Cloud API.

  • Next.js
  • pgvector
  • GPT-4o Vision
  • WhatsApp API
  • 110+ tests
• 04

  AEO Diagnostic · How AI assistants rank your brand

  LiveCode

  Fans buyer-style queries across 5 LLMs in parallel with per-cell timeouts, backoff and fallback, then extracts ranked brand mentions with structured outputs.

  • AI SDK
  • Upstash Redis
  • Structured outputs
  • 60+ tests
  • CI
• 05

  Healthcare Information Assistant · Grounded medical RAG

  Code

  Answers only from 9 whitelisted medical domains, scores each source's reliability with a separate LLM pass, cites inline and reads prescriptions with vision OCR. Temperature 0, no diagnosis.

  • LangChain
  • Tavily
  • GPT-4o Vision
  • Guardrails
• 06

  Nautilus · Multi-tenant store provisioning on Kubernetes

  Code

  Namespace-per-store Helm architecture with dedicated Postgres and Redis per tenant, a 10s reconcile loop, concurrency limits and timeout auto-recovery.

  • FastAPI
  • Kubernetes
  • Helm
  • k3s

More experiments on GitHub

1. Technical Consultant at Oorik Technologies

   Aug 2026 – Present
   • Built a zero-to-one hospitality ordering platform across kiosk software, hardware acceleration and backend services, from architecture through deployment.
   • Contributed to the AI Hub integration layer: OCR document scanning and LLM-agent workflows for quick-scan and financial decision support.
   • Grew the engagement to $5K MRR within the first six weeks.

2. Generative AI Intern at HYRGPT

   May 2025 – Jun 2026
   • Shipped an async AI phone-screening platform (FastAPI, LiveKit, Plivo) running live outbound interviews, with Redis token-bucket rate limiting and Postgres transcripts and verdicts.
   • Built an MCP server exposing a 13-step interview-scheduling workflow as tools, driven from Claude Desktop through a custom MCP client.
   • Orchestrated a 7-agent LangGraph scheduling system over Google Calendar, Gmail and WhatsApp, removing manual coordination for hiring managers.
   • Delivered a RAG pipeline for course discovery: Scrapy + Playwright crawling, overlap-aware chunking, Vertex AI embeddings and ChromaDB retrieval.

3. SDE Intern at KPMG

   Jun 2025 – Sep 2025
   • Built a meeting-intelligence bot that joins Google Meet and Teams, captures audio and processes meetings end to end. 1000+ meetings processed.
   • Designed a hybrid Whisper + Gemini pipeline for summaries, action items and sentiment at 99% uptime.

AI

LangGraph, LangChain, MCP, RAG, multi-agent systems, evals, guardrails, speech AI

Languages

Python, TypeScript, SQL, Go

Backend

FastAPI, Next.js, asyncio, Pydantic, Postgres, pgvector, Redis, MongoDB

Infra

Docker, Kubernetes, Helm, GCP, Vercel, GitHub Actions

• Manipal University Jaipur

  B.Tech, Computer and Communications Engineering

  2026

• IIT Madras

  BS, Data Science and Applications

  2027

Building something with LLMs? Let's make it actually ship.

Email is the fastest way to reach me.