Samuel Chin

Ph.D Candidate @ MIT

Cambridge, Massachusetts, United States

About

Experience

  • Graduate Research Assistant at MIT Center for Transportation & Logistics
    Feb 2022 - Present · 4 yrs 5 mos

  • Palantir Technologies (2 yrs 8 mos)
    • Software Engineer
      May 2019 - Jun 2020 · 1 yr 2 mos

      2X speedup of a PySpark application deployed globally.

    • Forward Deployed Software Engineer, Tech Lead
      Oct 2018 - May 2019 · 8 mos

      Upgrade and maintain applications in an air gapped Linux environment.

    • Forward Deployed Software Engineer
      Nov 2017 - Oct 2018 · 1 yr

      3X speedup of critical data pipeline in PySpark - 4K users in a global corporation used this application.

  • Research Assistant at Singapore University of Technology and Design (SUTD)
    Aug 2017 - Nov 2017 · 4 mos

    I worked in Dr. Sai-Kit Yeung's Vision, Graphics and Computational Design (VGD) Group. My research involved improving on existing image inpainting methods.

  • Engineer at D'Crypt Pte Ltd
    Mar 2017 - Jul 2017 · 5 mos

    Research Distributed Deep Learning on FPGAs. Discovered the non-deterministic behaviour in TensorFlow independently.

  • EFSG1 Cohort Entrepreneur at Entrepreneur First
    Sep 2016 - Mar 2017 · 7 mos

    Entrepreneur First selects the world's top technical individuals and supports them to build world-class tech companies from scratch. The inaugural cohort in Singapore attracted 2300 applications and I was part of the top 2.1% of candidates that were selected. I reached out to potential customers through cold emails, warm connections and anything that I could think of. Through the customer development meetings, I gained insights in many industries ranging from space to underwater. In the program, I went through 3 teams which all did not work out in the end and transited to consulting for some of the customers. The most notable thing I built was in the 3rd team. My co-founder and I built a 99.9% accurate single camera license plate detector within 2 weeks in TensorFlow. We did everything from scratch (except using pre-trained models of course) - collecting the data ourselves, preprocessing and storing it and finally training the models.