Mikhail Parakhin is a technology executive whose career includes search, machine learning, and large software organizations at Microsoft, Yandex, and Shopify. His recent public talks describe the organizational changes involved in using coding agents across an engineering team.[1][3]
Start with AI Stack Series: engineering organizations and autonomous AI →
Career
Shopify welcomed Parakhin as CTO in August 2024 to oversee engineering and data. Its announcement credits his Microsoft work on consumer and enterprise AI products, including Copilot, and his earlier role as Yandex CTO across search, cloud services, assistants, and ClickHouse. Those are leadership contributions to team-built systems, rather than sole-author claims. In October 2026, BetaKit reported direct confirmation from Shopify that he was leaving the CTO role and continuing as an advisor; the company did not disclose the effective date.[1][2]
Contributions
Parakhin also appears in OpenAI’s Codex Originals, a publisher-produced account of people using its coding tools. The segment places his Shopify work in the context of applying coding assistance at organizational scale. Taken together with the panel discussion, it provides an accessible account of how an executive thinks about adopting agents inside an established business, while keeping product promotion distinct from independent performance evidence.[4]
Ideas
In NEA’s AI Stack Series, Parakhin joined Eno Reyes and Mat Velloso to discuss engineering organizations built around agents. The publisher frames the shift as moving beyond faster text entry toward coordinating longer-running work. That raises practical questions about how work is specified, checked, and handed between people and tools. The discussion is useful as an engineering-leadership account of those changing responsibilities, rather than a benchmark proving that one agent setup is universally more productive.[3]
Tools, writing & talks
Writing, research & talks
Start here
A panel discussion with Parakhin, Eno Reyes, and Mat Velloso about agent-centered engineering work.
Start with the panel summary to see which team processes change when agents take longer assignments. No machine-learning background is required; experience reviewing software work helps.
Panel participant[3]Shopify's CTO discusses unlimited token access for all engineers, defaulting to the largest model, a digital twin of the company for predicting merchant growth, and a barbell token strategy for getting the most from Claude.
Catalog workOriginal source- AI-Native Engineering: 100% adoption, 5x search throughput, unlimited tokens — Mikhail Parakhinyoutube.com
Mikhail Parakhin describes trial-and-error findings on token spend, why parallel non-communicating agents waste tokens, and how multi-model critique loops raise code quality at the cost of latency.
Catalog workOriginal source - Codex Originals: Mikhail Parakhinopenai.com
OpenAI’s portrait of Parakhin’s use of coding agents at Shopify.
Featured interview subject[4]