# Pavan Raheja — pavan.blog

> AI Product Manager and builder. Builds AI agents and LLM products — with the governance, evals, and human-in-the-loop that make them work in production. Based in Dubai, UAE.

## About

Pavan Raheja is an AI Product Manager who ships: he built an evaluation-and-promotion pipeline that tested 60+ automated agent strategies and promoted only the 4 that cleared statistical rigor gates (a >90% kill rate), shipped Google Vertex AI to production at Mashkor (+15% activations, 3× MAU), and authored RFC #7218 on preventing catastrophic actions by autonomous AI agents. He works daily in Claude Code, MCP servers, and the Anthropic Agent SDK, and angel invests in AI and Web3 companies (xAI, GrowthX, WorldMobile, Worldcoin).

Core mantra: "Learn daily, grow daily, learn fundamentally."

## Key Experience

- Independent — AI Product Manager & Builder (Dubai, 2026–present): ships a fleet of production AI agents (autonomous + human-in-the-loop) running live capital across 4 venues; evals-driven — >90% of candidates killed before launch, out-of-sample validation, weekly live reconciliation, zero catastrophic incidents; full LLM stack daily (Claude API, MCP, RAG, evals); open RFC #7218
- Founder — Product & Operations at Dash Capital (Dubai): built the AI ops product that ran the business; 0 → AED 2M revenue in 18 months with full P&L
- Senior Product Manager (Growth & AI) at Mashkor (Kuwait): 3× MAU, 2.8× revenue in 18 months; shipped a Vertex AI recommendation engine (+15% activations)
- Growth & Product Consultant at PriceLabs (AI SaaS, 2020–21): growth/product partner through its 5× scale-up (team 7 → 150+)
- Growth & Product Lead at Flint Technology (boutique consultancy, 2011–20): client growth work — marketplace listings +240%, content platforms to 300K users
- Started as a software engineer (Java, microservices, banking applications)

## Writing

Technical writing on production AI agents — reliability, guardrails, human-in-the-loop, and evals — plus product management and growth frameworks.
Notable: "What a Year of Running Production AI Agents Taught Me About Reliability", "Risk Guardian: Preventing Catastrophic Actions in Long-Running AI Agents (RFC #7218)", "Building a Guardrailed AI Agent with Human-in-the-Loop".
Published at https://pavan.blog/articles

## Digital Clone

An AI-powered digital clone of Pavan is available at https://pavan.blog — built on the Claude API, trained on his knowledge, voice, and experience. Ask it about AI agents, product management, or his work.

## Contact

LinkedIn: https://www.linkedin.com/in/pavanraheja
GitHub: https://github.com/pavanraheja