ProfitPeak — Staff EngineerSep 2025 – PresentEcommerce profit analytics. I started the platform the product runs on — TypeScript, PostgreSQL and ClickHouse on AWS — and lead its development.
- Turned analytics from something customers requested into something they perform: they ask questions of their own data in plain language and get answers, instead of waiting on an analyst.
- Put that same capability inside the AI assistants customers already use, over MCP, so our analytics show up where they work rather than only in our app.
- Made the platform safe to sell upmarket — tenant isolation enforced at the database, role-based access across every surface, and access changes rehearsed against production data before they ship rather than discovered by customers afterwards.
- Replaced bespoke reporting with dashboards customers configure themselves, including forecasting that tells them what to do next rather than what already happened.
- Connected the systems customers actually run their business on — Shopify, Meta, Klaviyo, GA4 — so the numbers are complete enough to act on.
Dovetail — Staff Engineer & Search LeadJul 2023 – Sep 2025Led search, and the shift from finding documents to answering questions.
- Made generative AI a core part of the product's search rather than a feature bolted onto it, so users get answers with evidence instead of a list of results to read through.
- Let people interrogate their research conversationally — follow-up questions, in context, without learning query syntax.
- Kept it fast and affordable as it scaled, so relevance improved without the infrastructure bill following it.
MILKRUN — Senior Software EngineerJun 2022 – Apr 2023Full-stack engineer on the platform team, in a company growing faster than its codebase. I was laid off because the company shut down.
- Consolidated a fragmented estate of projects, stacks and release processes into one platform, making consistency and quality something the whole company could fix at once instead of team by team.
- Gave every service a shared, typesafe foundation, so correctness became the default and whole categories of bug stopped reaching production.
- Built the internal tooling the operations, support and content teams ran on daily, replacing manual work with a single interface and fine-grained permissions.
- Consolidated data scattered across Firebase, MongoDB and others onto PostgreSQL and PostGIS, which made the product both faster and easier to reason about.
IBM — Senior Software EngineerSep 2021 – Jun 2022I joined Envizi, which IBM acquired a few months later — not what I signed up for.
- Delivered an enterprise application that shipped and stuck, in TypeScript and C# / ASP.NET Core on SQL Server.
- Modernised the stack and the processes around it while the acquisition happened around us.
Yandex N.V. — Software EngineerApr 2018 – Mar 2021Worked on the platform for building skills for Alice, Yandex's voice assistant, used by millions.
- Rewrote the high-load core service from C++ to Java, cutting feature time-to-market from three months to one week — the platform could finally ship at the speed the product needed.
- Built the skill store that became the front door for the ecosystem, reaching 100,000+ daily active users.
- Made the platform reliable enough to stop being the reason things broke, with monitoring across 100+ business and technical metrics.
Mail.Ru Group — Software EngineerApr 2013 – Jul 2016Worked on ok.ru, one of the largest social networks in Russia, with 140+ million registered users.
- Replaced manual monitoring with a system analysing 5,000 charts every five minutes instead of 50, removing the human factor and giving the monitoring team their time back.
- Built the moderation tooling that cut the manual review load dramatically.
