Sep 2024 – Aug 2025 · 11 mos · United States

• https://cef.ai: Decentralized agentic workflows for Ambient AI, at scale.

Nov 2021 – Jul 2024 · 2 yrs 8 mos

• https://miro.com : The most valuable massively multiplayer online game ever created.
• End-user authorization, rules, evaluations at edge, performance & scale.
• Real-time policy evaluation with Open Policy Agent (OPA). Transpiled OPA policies to JVM.
• The DAG of data dependencies, OLTP pipelines, SAGAs, consistency models.
• Materialized views, CDC, Event Sourcing.
• Designed the shared-screen-first coding interview process to replace HackerRank.

Nov 2020 – Present · 5 yrs 6 mos

• https://github.com/SysDesignMeetup/sdm: We gather and talk system design, in a colloquium style.
• The channel got big since there's not much good SysDesign content for interview prep.
• The long play is to grow into a crowdsourced Architects Corner, where we can discuss real-life and artificial problems within the community, for fun and education purposes, as well as to bring value.
• On-air, recorded events are me or the community presenting on various topics.
• There is a lot more stuff going on behind the scenes. Come join us on Discord and Slack!

Aug 2019 – Sep 2021 · 2 yrs 1 mo

• https://www.pokerstars.com : The largest online poker in the world, acquired by Flutter in 2020.
• Architecting and sometimes helping implement, test, and maintain real-time data pipelines.
• Streamlining the procedures behind ongoing and upcoming regulatory reports.
• Led the automated cross-component testing initiative to run Dockerized tests natively from C++.
• Helping slice and dice and present product usage data to BI and product analysts teams.

Oct 2017 – Jan 2019 · 1 yr 3 mos

• https://friendlydata.io : Helping businesses make data accessible.
• Acquired by ServiceNow, which I joined as IC6 = L8, and left in a few months after rewriting our C++ into Java. Years later folks thanked me for clean and useful code.
• All things machine learning, query suggestions, spelling correction, backend for UX interactivity.
• Using labeled data to constantly improve the quality of the product in a semi-automatic fashion.
• Infrastructure, interactivity, reliability, repeatability, and overall data engineering.

Sep 2015 – Sep 2017 · 2 yrs · San Francisco, CA

• https://realtyshares.com : The Leading Online Marketplace for Real Estate Investing
• Data Consolidation:
• Single source of truth.
• Bi-directional sync between our proprietary data storage and Salesforce.
• Infrastructure:
• Zero-downtime rollouts of the data stack, including schema evolution.
• Fleet monitoring with dependencies visualization and sophisticated alerts.
• No data stack outages for 2+ years I was there, and for years later.
• Came back briefly in 2019 to perform knowledge transfer as the company got acquired.
• Leadership:
• Helped grow the engineering team to 10+, referred four people who successfully joined.

Jan 2014 – Present · 12 yrs 4 mos

• https://github.com/C5T/Current : Realtime machine learning, the C++ framework to develop and deploy data- and OLTP-heavy apps.
• Architecture:
• Publish-subscribe as the source of truth, now known as Event Sourcing.
• Leader-follower replication, eventual consistency between replicas.
• Zero-downtime data authority flip, with optional live schema evolution.
• Implementation:
• 10K+ HTTP QPS on one core.
• Convex optimization with seamless analytical differentiation.
• 100+K LOC, mostly infra, with 90+% test coverage.
• Battle-tested:
• Current is the backend of two production systems, one up and running to this day.
• And Current powered three proprietary data crunching / research projects.

Jan 2013 – Jan 2014 · 1 yr · San Francisco, CA

• http://www.staance.com : Opinions network. Enable people take stances and engage in conversations starting from a simple "Agree/Disagree" on popular trending topics.
• First hired engineer. Worked with the founding team.
• Introduced metrics for user engagement, ads efficiency, behavioral insights.
• Grew into an engineering leader, Oversaw data science and product development.

Jan 2012 – Jan 2013 · 1 yr · Bellevue, WA

• https://angel.co/public-verification/ : Personal identification in under one minute. AirBnb- and Uber-like reputation for everyone, everywhere.
• Built and led the team of four engineers.
• Define the products and the APIs.
• Work with investors and external stakeholders.
• Integration work, both as the provider and as the consumer of the service.
• Provisional patents:
• "User Identity Verification Via Credit Card Information."
• "User Identity Via Mobile Technology."

Jan 2011 – Jan 2014 · 3 yrs

• http://www.maps.me : Offline maps for mobile devices based on rich OpenStreetMap data.
• Now known as Organic Maps, which I still use often.
• Business networking.
• Data analytics.
• Technology stack.

Jan 2011 – Jan 2012 · 1 yr · Bellevue, WA

• http://bing.com/
• Airfare prediction: http://www.bing.com/travel
• Image content features understanding: http://www.bing.com/images
• Vertical search ranking and more machine learning projects.
• Patent:
• "Determination of Relationships Between Collections of Disparate Media Types."
• Talk:
• "Gödel Theorems for engineers and computer scientists".

Jan 2010 – Jan 2010 · 0 mo · Cairo, Egypt

• http://www.guc.edu.eg/ : The German University in Cairo (GUC).
• Coached the teams for the ACM ICPC competition.
• Co-taught the “Data Mining” class.
• Talk:
• “Machine Learning: New Horizons it Opens for Web Applications”, opening Cairo IEEE Day.

Jan 2007 – Jan 2011 · 4 yrs · Zurich, Switzerland

• http://www.google.com
• Search quality: Unwanted explicit content demotion in search results.
• GMail: Site reliability engineer.
• Infrastructure work: Scaling spam and abuse pipelines for GMail, used for Google+ later on.
• Patents:
• "Classifying documents using multiple classifiers."
• "Construction of Text Classifiers."
• Talks:
• One on scaling the spam/abuse pipelines.
• One on engineering productivity.
• Orientation class for new hires:
• Crafted and ran a workshop on how to work with Google codebase.
• Kicked off in Zurich, got broader adoption soon.

Jan 2006 – Jan 2007 · 1 yr

• Non-relational quadruplet-based storage system, with PM-friendly UX.
• Core DSL to insert, update, and query data.

Jan 2005 – Jan 2006 · 1 yr

• Text mining framework with user-friendly UX.
• Supervised and unsupervised text categorization.
• Compression, by itself and as one of text similarity metrics.
• SVM and entropy models, text stemming, and a few other in-house algorithms.