Product manager · 20 years shipping technology products · currently building AI products.

I've led product development at robotics startup (outdoor autonomous robots — SLAM, CV-ML, ROS, 35+ engineers, first sale), run a decade as an independent consultant (helped one client grow from ₽800M to ₽4B), and launched public-service organization. Today I'm focused on the intersection of product management and applied AI: prompting, RAG, agents, MCP, evals.

Three small, deliberately-scoped pet-projects exploring what AI-native knowledge work and agents actually look like in 2026:

• study-vault-template — a template Obsidian vault where Claude (or any capable LLM) acts as the wiki's operator. The pattern: LLM as compiler of persistent wikis, not RAG retriever. Includes a ~400-line CLAUDE.md operator manual.
• airesearch — a daily scout that watches six AI blogs, scores each article against my current research topics with Claude Haiku, and writes a dated digest into an Obsidian vault. Claude compiles anything I actually read into a living wiki.
• aibirthday — a five-agent pipeline that drafts, approves, and sends personalised birthday greetings via Gmail + Claude Opus with adaptive thinking. Consent gate is a tag in the contact's notes, not a daily confirmation — a small detail that makes the thing actually usable.

Common thread: the operator manual (CLAUDE.md / prompts / YAML schemas) is the product; the code is scaffolding. This is what a spec looks like when the implementer is an LLM.

• Chief product manager at 168robotics (outdoor autonomous cleaners) — 2023–2025
• Independent consultant / interim CPO — 2014–2023
• Project and product lead on the MFC project for the Republic of Sakha (Yakutia) — 2009–2011. Awarded by the President of Sakha for socio-economic contribution.
• NES (MAE'14) · OmSU (Economist, 2004)

Senior / Lead / CPO roles in AI-native or robotics product work. Moscow-based; open to remote or relocation. Happy to work with foreign teams.

Reach me: akireykov@nes.ru · Telegram @andreynki