# bharatagarwal.dev — Full Context

> Bharat Agarwal is an AI Infrastructure Engineer based in Gurugram, India. He builds agentic systems, MCP servers, LLM evaluation pipelines, and multi-agent orchestration infrastructure. This is his professional site and builder's log.

## About Bharat Agarwal

Bharat Agarwal runs a GST-registered sole proprietorship (UDYAM-HR-05-0180458, Micro Enterprise, NIC 62020 — Computer Consultancy) focused on AI infrastructure consulting. He has previously worked at Shopify (Kubernetes autoscaling to 2.4M cores for BFCM), Atlassian (chaos engineering, incident response, co-inventor on US Patent US20220207086A1 for graph-based collaborative document interfaces in Confluence), and studied computer science at Bradfield School of Computer Science (networking, memory management, concurrency).

Before software, he completed five years of architectural training at SPA Delhi (School of Planning and Architecture). He also co-created Mothership, an open-source PaaS.

### Current work

- Agentic systems architecture: knowledge graphs (5000+ nodes mapping curriculum dependencies), intelligent agents, multilingual video processing pipelines (24hrs to 3min)
- Infrastructure at scale: Kubernetes autoscaling to 2.4M cores / 4.7M req/sec (Shopify BFCM), API redesigns cutting response times from 30s to 9s
- LLM evaluation pipelines: deterministic eval frameworks wired into CI/CD, hallucination detection, prompt regression testing
- Multi-agent orchestration: MCP servers, cognitive routing (brawn tasks to Challenger APIs, brain tasks to frontier models), durable execution
- Agent coordination tooling: Beads for agent memory and task coordination, Gas Town patterns for multi-agent CLI orchestration (10-30 parallel agents sharing a task graph)

### Management philosophy

Every team member guards one zone — UX, metrics, evals, architecture. You can write anywhere, but you only guard one place. Borrowed from Bill Belichick's "Do Your Job" principle and applied to agentic pods. Works the same whether the teammate is human or an AI agent.

Building toward two patterns: ephemeral workers (session-scoped agents that spawn, execute, and dissolve) and deployed workers (persistent agents wired into infrastructure, running eval loops and monitoring pipelines without human intervention).

## Key Content

### An MCP Server Gamma Should Have Built
URL: https://bharatagarwal.dev/gamma-mcp.html
Published: 2026-02-28
Source code: https://github.com/bharatagarwal/gamma-mcp-server

Bharat built a custom MCP (Model Context Protocol) server for Gamma's presentation API because Gamma does not offer a public MCP server despite having a REST API. The server connects Claude Code to Gamma, enabling presentation generation from the terminal.

The architecture evolved through three transport stages: (1) stdio — MCP's default, one client per session; (2) HTTP via StreamableHTTPServerTransport on port 24748, supporting multiple clients with UUID-based session management; (3) macOS launchd agent for persistence across reboots.

Five tools: generate-gamma, generate-from-template, list-themes, list-folders, check-generation-status. Built with the Model Context Protocol TypeScript SDK. ~400 lines of TypeScript across 8 files.

### GEO: Making a Site Citable by AI
URL: https://bharatagarwal.dev/geo-for-llms.html
Published: 2026-03-01

Documents the implementation of Generative Engine Optimization on bharatagarwal.dev. Covers the GEO playbook (structure for extractability, llms.txt standard, JSON-LD schema markup, E-E-A-T authority signals, semantic richness) and the exact implementation applied to this site.

References the Princeton KDD 2024 paper (arXiv:2311.09735) that formalised GEO, and Jeremy Howard's llms.txt proposal from Answer.AI (https://llmstxt.org/).

## External Profiles

- LinkedIn: https://www.linkedin.com/in/bharat--agarwal/
- GitHub: https://github.com/bharatagarwal
- X/Twitter: https://x.com/bharatagarwal__/
- Personal blog: https://bharatagarwal.io

## Entity Details

- Legal Name: Bharat Agarwal (Sole Proprietorship)
- GSTIN: 06BQWPA8293E1ZG
- UDYAM: UDYAM-HR-05-0180458 (Micro Enterprise)
- Address: Gurugram, Haryana, India
- Email: bharat@bharatagarwal.dev

## Verifiable References

- Shopify BFCM infrastructure: https://shopify.engineering/bfcm-readiness-2025
- Atlassian patent (co-inventor): https://patents.google.com/patent/US20220207086A1/en
- Bradfield School of Computer Science: https://bradfieldcs.com
- Mothership PaaS: https://mothership.live
- Gamma MCP server source: https://github.com/bharatagarwal/gamma-mcp-server
- GEO academic paper: https://arxiv.org/abs/2311.09735
- llms.txt specification: https://llmstxt.org/
- MCP TypeScript SDK: https://github.com/modelcontextprotocol/typescript-sdk