Jun 2023 – Aug 2025 · 2 yrs 2 mos · Zurich, Switzerland

• Cutting edge AI/ML for Siri and Apple web search.

Aug 2022 – Jun 2023 · 10 mos

• Cloud Conversational AI
• I deliver AI based conversational products to Google Cloud customers.
• 1. Smart Reply solution: a LLM-based system to to suggest appropriate replies to agents that converse with customers (e.g. tech support agents), scaled to 5 language.
• 2. Summarisation solution: an LLM-based to summarise conversations between agents and customers, scaled to 4 languages

Aug 2021 – Aug 2022 · 1 yr

• As a Tech Lead of the User Language Profile team of 26 people I am responsible for delivering content to users in the right language. The ULP product is used by most Google services (Search, Search Ads, Display Ads, Youtube, Discover, etc) to understand in what languages to serve content to their users.
• I led several efforts to improve Google products with better language targeting using ULP:
• ● Delivering >$XXXM ARR (+1.3% of the network-wide revenue) with better language targeting in Display Ads
• ● Delivering >$XXM ARR with better language targeting in Search Ads
• ● Helping Search to make information more accessible to Ukrainian users in 2022
• ULP is a ML-based system for predicting in which languages Google can serve content to the particular user. I led several efforts to improve ULP itself:
• ● Designed a cookieless ULP model that solves the "cold start" problem of predicting lanuages for new users
• ● Designed a global NN-based ULP model (instead of per-country linear models), delivering +5% prediction quality
• ● Proposed a new transformer-based ULP model, further improving prediction quality and showing extra >$XXXM ARR in Display Ads in early experiments

May 2020 – Aug 2021 · 1 yr 3 mos

• Co-founder and Tech Lead of the Content Recommendation.
• Content Recommendation is a search engine to recommend content to users of Goole Display Ads publishers. The product delivers O($XXXM ARR) and is used by ~180K publishers.
• I re-established the team in London after it got reorged-out in Mowntain View with a team of 3 engineers rebuilt most of the search engine that was designed for 100x less scale than I grew to be. The new stack processes O(10-100 PiB) of data to refresh the index daily, which at the time of the reorg was stale for several months.

Nov 2017 – Aug 2021 · 3 yrs 9 mos

• Co-founder and Tech Lead of the Publisher Brand Safety effort in Google Display Ads network.
• Publisher Brand Safety is an x-org effort across Ads and Trust & Safety to focus on the Brand Safety of Google Display Ads publishers and long-term inentory retention in GDA.
• Delivered a x-org pipeline and a ML system allowing Trust & Safety org to utilise la large volume of crowdsourced publisher blocking decisions to detect and ban policy violating advertisements. Resulted in cumulative ~300x YoY improvement in preventing policy violating impressions, O(10B) monthly.

Feb 2016 – May 2020 · 4 yrs 3 mos

• Ad Review Centre (ARC) is a search engine, where Google Display Network publishers can review, search and configure ads Google serves on their properties in order to maintain brand safety with Google Display Ads.
• I led a team of up to 13 people (up to 8 engineers), across US and Europe, to rebuild multiple aspects of the product.
• ML-based ranking system: I built an ML-based ranking engine for the search engine, from scratch, allowing publishers to be more efficient in fighting unwanted ad impressions, making it prevent ~12x more bad impressions than previously.
• Near-real-time indexing: delivered an infrastructure redesign to fully update a search index with ~16B documents daily, ~6x improvement from 20% of documents indexed daily to 100% with an extra margin for future growth. Indexing systems go thourgh O(10-100 PiB) of data daily.
• Multimedia search: built capabilities to search ads by images, both assets and screenshots.

Nov 2014 – Nov 2020 · 6 yrs

• As a co-founder and CTO I had founded and grown the engineering organisation from just myself to 16 people, including Software, Hardware and Research engineers and medical scientists. I led the team to deliver our AURA Band and AURA Strap products, from initial conception to market.

Jan 2014 – Feb 2016 · 2 yrs 1 mo

• At Yandex I was responsible for improving the search engine ranking quaity through improving ML-based ranking models. My innovations in ranking features, had cumulative impact of ~1% on search quality. My tooling, built to scale contribution of the rest of the team, led to anoter ~1% extra cumulative gain.

Oct 2012 – Sep 2013 · 11 mos · Moscow

• Data analysys and visualization, web frontend development. Also a little backend.

Apr 2012 – Apr 2013 · 1 yr · Moscow

• Game development for Android and PC.