Challenge
Cloud.ru's supercomputer is expensive, high-demand infrastructure used by large ML teams. The problem wasn't just running tasks (you can do that in Jupyter) — teams needed to manage access, coordinate resources, assign roles, and administer collaborative workflows at scale. Without a unified platform, organizations couldn't efficiently control who accesses what, track resource usage, or manage team permissions.
Role & Timeline
Product Design Lead
Led a team of 4 designers for 2 months (MVP), then continued as design director on the client side for 2 years.
Approach
Phase 1: MVP in 8 weeks
Built core design system: tables, filters, cards, states
Designed team administration: user roles, permissions, project access
Created unified navigation connecting all services (Jupyter, Jobs, Data Transfer, Pipelines)
Established patterns for resource monitoring and task management across teams
Phase 2: Growth & optimization
Analyzed behavior of ~20 corporate teams using the platform
Improved collaboration scenarios: role assignment, resource allocation, team coordination
Iterated based on real user feedback and usage analytics
Results
+40 NPS — high user satisfaction
Team administration — centralized control over roles, access, and resources
Cost efficiency — better resource management for expensive supercomputer infrastructure
Scalable platform — actively used by dozens of ML teams on Cloud.ru