Experience

Industry

Amazon | Seattle, WA → New York, NY
Applied Scientist, May 2021 - present
Research Engineer, July 2020 - May 2021

  • Trained 1B param multimodal foundation model with large-scale vision-language-audio pretraining. Out- performs OpenAI CLIP by 25% on internal zero-shot classification and retrieval benchmarks.
  • Enabled automated video advertisement insertion (CEO-level goal) with novel video segmentation model.
  • Developed embeddings for visual search and recommendation which outperform baseline recsys by 5%.
  • Trained multimodal transformers for automated content moderation and compliance.
  • Built distributed PyTorch training codebase and managed compute infrastructure for larger org of 30+ ICs.
  • [ICCV 2023 (1st auth)] Improves SOTA video masked autoencoders by 5% in action recognition. Amazon blog.
  • [ICLR 2023 (1st auth)] Nearest-neighbor sampling improves positive pair diversity for video contrastive learning.
  • [CVPR 2021] Self-supervised learning improves state-of-the-art movie segmentation by 13% while reducing annotation by 75% (saving $200K/yr) and speeding up training by 84%. Amazon blog.
  • Mentored two research interns to full-time offer.

Amazon Web Services (AI) | Seattle, WA
Software Engineer, August 2019 - July 2020

  • Elastic Inference reduces inference costs by enabling users to flexibly provision GPU compute.
  • Added logging metrics and launched canaries to support new EC2 G4 instance family.
  • Launched Elastic Inference-enabled PyTorch framework for SageMaker, EC2, and ECS (see my AWS blog post). Implemented TorchScript graph validation, shipped updated AWS Deep Learning Conda environments and Docker containers, benchmarked performance for vision and NLP models on multiple platforms, and wrote technical blog post.
  • Created proof-of-concept for building and integrating TensorRT-enabled TensorFlow 2.1 into the inference engine. Reduced latency by up to 70% compared to FP32 native TensorFlow in benchmarks.

Amazon Web Services (Databases) | East Palo Alto, CA
Software Engineering Intern, June - August 2018

  • Developed automated devops tool for AWS Aurora, a distributed cloud-native relational database, which improved on-call engineer productivity by automatically applying fixes to low-severity tickets and reducing manual processes.
  • Wrote a tool for enabling/disabling autoscaling policies and provisioning IOPS on DynamoDB clusters to improve cost management.

Phosphorus | New York, NY
Software Engineering Intern, May - August 2017

  • One of 30 Princeton Start-up Immersion Program participants placed at early-stage or series A startups.
  • Redesigned and implemented custom UI/UX components for user dashboard using Wicket and Scala.
  • Designed modeling layer in Scala, Spring Boot, Hibernate, and PostgreSQL. Wrote AWS CloudFormation templates for automated infrastructure deployment.

Academic/Research

Princeton Vision and Learning Lab | Princeton, NJ
Undergrad Researcher, September 2018 - current

  • Did an undergraduate senior thesis in computer vision and deep learning, related to single-view 3D vision.
  • Paper accepted to CVPR 2020 conference.

Harvard-MIT HST | Boston, MA
Bioinformatics Research Intern, June - August 2016

  1. Claims-wide Association Study (CWAS)

    • Worked with Prof. Isaac Kohane and Dr. Arjun Manrai to develope a new association study called “claims-wide association study (CWAS)” - like genome-wide association studies (GWAS), but for insurance claims.
    • Built a data visualization tool for plotting heatmaps of the USA from parsed AETNA insurance claims, at multiple levels of geographic specificity (zipcode, county, state, regional)
    • Used R, MySQL, and the Shiny web framework; code here.

  2. User-friendly Bioinformatics Tools (UBiT2)

    • Worked with Dr. Jean Fan to develop an open-source web application for client-side RNA-seq and qPCR data analysis.
    • All computation and data visualization is done client-side, thus providing a secure and fast environment for bioinformatics that involves no server.
    • Built in HTML, CSS, and Javascript. Code here.
    • Technical report posted on bioRxiv.

Rutgers New Jersey Medical School | Newark, NJ
HS Research Intern, June - August 2014

  • Investigated effects of the Ras•GTP-Raf-MEK-ERK signaling pathway in Drosophila fruit flies on organismal and organ senescence.
  • Conducted lifespan and stress (starvation, oxidative and heat) assays on flies with transgenes expressing varied levels of Rpd3 protein in the heart tissue.
  • Conducted heartbeat measurements on flies throughout various stages in their lifespan to analyze heart-function decline with age.
  • Co-authorship in a paper published in Aging, and semifinalist status in the 2015 Intel Science Talent Search (STS).

Rutgers Dept. of Physics | Piscataway, NJ
HS Research Intern, June - August 2013

  • Explored symmetry-breaking phase transitions in multiferroic materials (i.e. rare-earth hexagonal manganites).
  • Polished and imaged these materials.
  • Analyzed topological defect distribution using published theoretical results.
  • Co-authorship in a paper published in Nature Physics.