panel discussion · AI Dev Conf 2026 · AI4SDLC
AI Yesterday, Today, Tomorrow
May 21, 2026 · AI Dev Conf 2026
An AI Dev Conf panel discussion on how AI in software development has become baseline engineering hygiene and what it takes to embed it safely into a production process. The conversation covers context, agentic coding, metrics, policies, model routing, and the shift from local acceleration to measurable outcomes.
Panelists
- Alexey Totmakov · VK Tech
- Rafael Tonakanyan · Sber
- Andrey Kuleshov · Yandex SourceCraft
- Alexander Polomodov · moderator
Discussion focus
- 01Why is a model without context similar to a confident junior engineer, and which context does it need to be useful?
- 02What changes when teams move from AI assistants to agents that plan, edit code, run tests, and retry?
- 03Which access boundaries, approvals, sandboxes, rollbacks, and security checks are needed for agentic development in a large company?
- 04Which metrics show production impact rather than adoption alone: cycle time, time-to-PR, review quality, defects, risks, and cost?
- 05Why should rules for working with code, logs, documentation, and internal data become machine-checkable policies?
- 06How can teams build a strategy around gateways, routing, evals, observability, cost limits, and provider replaceability instead of one model?