Experienced Principal PM at Microsoft, deliberately rebuilding myself as an AI-native PM — not by reading about agents, but by shipping with them.
What that looks like in practice
I've written a growing library of agent skills — packaged, config-driven workflows for coaching, strategy work, presentations, and personal automation. The first one going public: ai-pm-course, an AI PM interview-prep coaching course that runs as an agent skill — chapter → quiz → graded drills → progress tracking, from a 243-question bank and 34-chapter curriculum, built so anyone can run it with their own question bank. Coming soon.
I also build products with coding agents (Claude Code, Codex) doing the implementation while I own product boundaries, decisions, and approvals. Every repo carries an AGENTS.md with operating rules and approval gates, so any agent can resume the work cold. The bet: the PMs who matter next are the ones who can direct AI systems, not just spec for humans.
Products (in progress)
- Retirement Planner — Goal-based financial planning for Indian households. Deterministic Python computes every number; the LLM only explains, proposes, and challenges — it never produces a figure the plan depends on.
- Entry Desk — A human-gated decision desk for NSE equities. A versioned rulebook maps frozen metric snapshots to decisions; guardrails can force abstention. Adopted rules, not forecasts.
Find me — LinkedIn