Software Engineer - ML, Google · San Francisco

Abdul Hannan Kanji

I build machine learning models for ads at Google.

Before that: reinforcement learning and ML-safety research at UMass Amherst, open source with Microsoft’s Fairlearn, and four years of full stack & mobile work at Intuit.

Abdul smiling, with curly black hair and a beard, in a blue T-shirt against a plain wall

At Google

Since 2022

Software Engineer - ML · Ads

I work on the machine learning behind ads on Google, taking ideas from early analysis and design through to production, together with modeling, serving, and product teams.

Before Google

  • SwayAIApplied Data Scientist2021–22

    Enterprise ML on AWS SageMaker: visual anomaly detection for an aeronautics client, object-detection serving for radiology, and call sentiment analysis, rebuilt as parts of a no-code ML platform.

  • Reich Lab, UMass AmherstResearch Engineer2020–21

    Kept the COVID-19 Forecast Hub’s data pipeline running, the official data source for the CDC’s forecasting page, and cut submission validation from about 20 minutes to under one for 50+ research teams.

  • IntuitSoftware Development Engineer2015–19

    Built the Android app shell that QuickBooks, TurboTax, and Mint mobile shared for auth and logging; shipped QuickBooks transactions features and on-device ML receipt validation.

Research & open source

Google Scholar

  • Seldonian safetywith Prof. Philip Thomas, UMass Amherst · 2020–21

    Training models that come with a high-confidence safety guarantee (Thomas et al., Science 2019). I added stratified sampling, which raised the chance of returning a safe solution by about 60% in worst-case classification settings, and built safe policy improvement for RL on per-decision importance sampling.

    DocsCode

  • Fairlearnwith Microsoft’s Responsible AI team · 2020

    Implemented the Equality of Opportunity fairness constraint and made ExponentiatedGradient’s per-iteration estimator copying cheaper. Both merged.

    Equality of OpportunityExponentiatedGradient

M.S. Computer Science, UMass Amherst 2019–21B.E. Computer Science, PES Institute of Technology 2011–15

Off the clock

Outside work I hike, run, and play the drums.