Releases and cross-team delivery for an 80-person engineering organisation.

• 2026
• Geo-tech platform
• Rolled out to the team

A delivery system for feature owners

A feature owner gets asked three questions all day — what's the status, will it slip, why did we decide it that way. With five features in flight and a dozen more in support, the answers stop fitting in one head. So I built somewhere for them to live. Since September it runs for the whole team: every epic of the quarter, across all feature teams, sits on one managers' dashboard that is published inside the company and rebuilt every morning.

• Spec, decision log and risk register with owners and review dates; closed risks keep their outcome instead of disappearing
• Daily status snapshots pulled from the tracker, append-only — yesterday can't be tidied up today
• Forecasting that reads closing speed and scope growth together, calibrated on the feature's own history
• Documentation and tickets updated automatically after discussions; a release dashboard rebuilt every morning

Read the case →

• 2026
• Own product
• Live beta

Grumli — find the gaps before your team does

Hand a vague spec to a developer and you get a question. Grumli asks those questions first: paste a brief, a spec or a user story and get the document back completed, with everything that would have surfaced at grooming already marked up.

• Ambiguities quoted, explained and rewritten against EARS and IEEE 830
• Missing cases — errors, empty states, permissions, boundaries, concurrent access, rollback
• Acceptance criteria you couldn't write a test against, flagged as such
• A traffic-light verdict: ready to hand over, send back to the author, or needs a stakeholder decision

About the product →

• Apr 2024 — now
• Geo-tech platform
• Remote

Release and delivery management

Since September 2026 I'm the single point of contact for releases, feature teams, the release calendar and status sync across an 80-person engineering organisation. I handle every epic of the quarter in Agile delivery, from planning through release, and own the team release end to end.

• 12+ cross-functional teams of 6-8 people, three to four active at a time
• Every epic of the quarter planned, tracked and driven to release, with Scrum and Kanban across teams
• The release calendar and a quarterly tracker where my date overrides the calendar; a release console for release day
• Architecture decisions driven together with the engineering team

• 2018 — 2024
• Cian · Yandex · Ozon
• Data

Clickstream, warehouse, pipelines

Where the technical half of my background comes from, and the reason I can follow an architecture discussion without a translator.

• An in-house user event tracking clickstream system at Cian, worth more than $160k a year in saved cost
• Data delivery and storage in a Data Vault warehouse at Yandex; space-cleansing cut cluster usage by 20%
• The Seller System platform at Ozon, taking recommended-item accuracy to 93%
• Documentation for microservices and APIs, reports and user statistics throughout

• 2026
• Open source
• MIT

amber: Markdown editing for self-hosted Confluence

Teams on Confluence Server and Data Center can't edit pages in their own editor, review changes as diffs, or let AI agents work with their wiki. amber turns a page into Markdown and brings it back without breaking macros, panels or layouts.

• Anything Markdown can't express is set aside before conversion and restored byte for byte
• A visible marker where each macro sits, so it can be moved on purpose and never lost by accident
• Refuses to publish over a colleague's edit made after your pull

See the code →