Aleksandr Korolev

Let the founder raise tariffs without losing the base — churn risk quantified per cohort before the price move, and managed after it.

Analytics layer built on proprietary billing, tickets and OpenStack metrics with AI-coding tools instead of queueing for engineering; forecast shipped as an operating tool with targeted-discount levers, not a report.

Product Owner reporting to the founder; solo, AI-assisted build — engineers stayed on infrastructure.

Stopped margin leaks — VPN/proxy abuse on unmetered plans was making whole regions unprofitable.

Detection balanced against unit economics (region revenue vs transit cost), not a blunt ban-hammer; tariff what-if on real reference prices.

Product Owner; designed and deployed end-to-end, AI-assisted.

Replaced expert memory and weeks of manual catalog work with reproducible cross-collection search that works from the field.

Visual-similarity search instead of a classifier — made and defended the key call; data pipeline with validation sets, expert-quorum review, drift control via cross-validation.

8 people — ML engineers, backend/frontend, archaeologists; sole Product Owner, full lifecycle.

Made the impossible catalog feasible — manual comparison against the British Museum's 26,000 prints would have taken one expert 20+ years.

CV pilot matched the two collections, then the British Museum's richer open data was adapted via LLM; a public catalog added to the PRD to secure funding and measurable scientific impact — defended to museum leadership.

CTO / Product Owner; dual-track kanban (scientific annotation separate from dev); ~20-intern program with access to closed Hermitage vaults; 58 curator queries as QA.

Turned grant-cycle research into shipped products, reusable datasets and a partner network that outlives any single grant.

Product-driven operating model for an academic lab; GPU-cluster design for inference; an original AI & prompt-engineering course and an AI-friendly university policy as the talent pipeline.

Co-founder & CTO; product teams up to 8 people; 50 interns a year.

Gave enterprise clients a compliant answer to shadow IT — corporate documents leaking into personal Dropbox and Google Drive accounts.

Dogfooding with 96 internal users as the main validation channel; 10 stable modules curated out of 200+ instead of shipping the raw open-source constructor; Pay-per-Resource billing, SPLA licensing, 152-FZ certification, Office 365 migration path.

Sole PM; 2 engineers, DBA, product sales, PMM, part-time S3 specialists, vendor engineering teams — ~10 stakeholders coordinated.

Made an abandoned, loss-risking product line profitable and competitive against faster-moving hyperscaler rivals.

P&L audit traced losses to VMware licensing driven by RAM over-allocation — the tariff grid was rebuilt around real utilization; HA Cluster and Disaster Recovery tiers unlocked the >$1,000-instance segment.

Sole PM; enlisted the monitoring team's data to win the C-level argument.

Kept the product inside tenders — regulated clients (Central Bank of Russia, Rosatom, energy, banks) could not wait two years for current hypervisor support.

Design-to-development framework — prototype, spec and technical risks in parallel instead of sequential handoffs; user stories with regulator-compliance acceptance criteria; UX testing introduced into an engineering-led culture.

Product Manager owning the full lifecycle; synchronized product design, UX and development.