Ben Sidhom
  • Software Engineer
  • Data Practitioner
  • Counting Enthusiast
Contact: <first>@<last>.io
Seattle, WA

Work experience

    • Software Engineer
      2017—
      Cloud Dataproc
      Google (Remote)
      • Designed, implemented, and benchmarked various experimental and production shuffle backends for Spark, including Dataproc Enhanced Flexibility Mode. (Tech lead.)
      • Worked with Spark community to get API changes into Spark core to support disaggregated shuffle implementations. (Tech lead.)
      • Worked on initial Native Query Execution engine for Spark on Dataproc.
      • Worked on Spark Connect/Python notebook client libraries and integration.
      • Designed and implemented the first Portable runners for Apache Beam (on Flink backend).
      • Identified and performance bugs in the GCS connector and Dataproc distribution to improve best-case throughput by 20% and worst-case (pathological) inputs by orders of magnitude. (Inclues JNI work.)
      • Designed and implemented core end-to-end/integration testing framework for Dataproc.
      • Specialized in Spark, performance, price-performance, and distributed reliability engineering.
    • Software Engineer
      2014—2016
      Android Machine Intelligence
      Google, Seattle
      • Adapted early FaceNet models to run locally on device.
      • Compiled and compressed knowledge graph models to run locally on devices. Included distributed processing pipeline and Android client work. Technology was integrated into GBoard and other core Android services.
      • Android application-level programming and JNI.
    • Software Engineer/Data Scientist
      2013—2014
      Shipping Science
      eBay, Bellevue
      • Added incremental improvements to production Fast 'N Free model, with a focus on latency and model accuracy.
      • Developed new extensible data transformation and training pipeline in Spark
      • Implemented Akka-based system to automate data generation and verification.
    • Software Engineer Intern
      2012
      Shipping Science
      eBay, Redmond
      • Designed and implemented new Fast 'N Free shipping estimate machine learning model.
      • Trained and tested model on Hadoop, outperformed then-current system.
      • Model was used on live site 2012—2013 (after internship).

Education

  • University of Washington, Seattle Graduated June 2013
  • Double Major: Bachelor of Science, cum laude, Computer Science and Physics
  • Minor: Math

Skills

  • Distributed systems
  • Machine learning
  • Bayesian probability modeling
  • Spark

Programming Languages (recent experience)

  • Python
  • Rust
  • JavaScript
  • Java

Programming Languages (professional experience, not recent)

  • C++
  • Scala

Miscellaneous Interests

  • Fully persistent data structures
  • Software security
  • The power of randomness
  • Programming languages
  • Skiing
  • Biking (road and XC mountain)
  • Hiking (goal to complete Washington PCT in individual day-segments)