Rolling out agents
Moving engineering to working with AI agents: Claude Code and Codex in daily tasks, working rules, review and acceptance of results.
CTO · 15+ years in software
I lead engineering and infrastructure and drive the AI transformation of the engineering team and the whole company.

Moving engineering to working with AI agents: Claude Code and Codex in daily tasks, working rules, review and acceptance of results.
Building analytics that shows what AI actually delivers: who uses agents and how, what it costs and where it pays off.
Connecting agents to work systems through MCP servers. The hub keeps them in one place, each with a ready-made prompt the agent uses to install it.
Teaching the team to work with agents: practices, task reviews and internal guides, so everyone uses AI with confidence, not just enthusiasts.
Novosibirsk
Notes (RU) →“Adopting AI and measuring results”
Saint Petersburg
Saint Petersburg
Notes (RU) →“Roles in the age of AI”
Saint Petersburg
Saint Petersburg
Notes (RU) →“How C-level one-on-ones work”
Moscow
Talked about the AI revolution in teams. The new baseline is telling an agent by voice to go through MCP and get the task done.
Moscow
Met Ozon's CTO and talked about adopting AI in engineering and whether to set targets for it.
Yandex and T-Bank offices, speed one-on-ones with the bank's leaders, talks on AI in engineering and how to measure its effect.
Four days in the mountains. About 20 new CTO acquaintances, swapping experience of adopting AI.
Timeweb Cloud business breakfast. The case is published on Timeweb's Habr blog.
Saint Petersburg
Earlier: Bitva za IT: the IT job market and hiring (panel, 2024) · HighLoad++ (2017, 2019, 2021–2024), TeamLeadConf (2019, 2021–2023), DevOpsConf (2021, 2022), SOUTH HUB (2024), #highload masterclass (2016)
I help engineers and new managers. Calls are booked via GetMentor.




An Android and iOS audiobook player on top of knigavuhe.org. Offline chapters, book series, sleep timer and a saved position for every book.
An offline Android voice recorder that uploads to Google Drive. New recordings are transcribed on a GPU, and AI turns the text into notes.
How to chain a phone, cloud storage and local transcription into one pipeline.
Voice input on a hotkey. The GigaAM-v3 model recognises speech locally on the GPU, and the text, already punctuated, is typed straight into the active window.
I compared T-one, Whisper and GigaAM on my own speech and kept GigaAM. It is faster on short phrases and adds punctuation by itself.
Converts typeset PDF books to FB2 and EPUB. There is a CLI with a Claude skill and an in-browser version that never uploads the file.
Keep the mechanics in code and leave the judgement to AI.