Samvit Jain
Software Engineer, ML at Google
Greater Seattle Area
Overview
Work Experience
Software Engineer, AI/ML
2022 - Current
AI/ML engineer on the Payments Risk team. As ML area owner, drove multi-year effort to overhaul Google’s approach to chargeback disputes: - Developed & launched models to (1) identify chargeback abuse and (2) estimate Google's likelihood of winning a dispute. Replaced broken legacy policies with a simple, maintainable ML-based system. - Impact: Recouped millions in GMV and fees savings for Google annually via new model launches and better chargeback defense decisions. - Contributions: Selected features, developed datasets, trained models, defined impact metrics, implemented policies, analyzed experiments, drove alignment, and shepherded new model launches.
Google is a multinational corporation that specializes in Internet-related services and organizes the world’s information.
Raised $26,100,000.00 from Angel Investors LP, Kleiner Perkins, Signatures Capital, Sequoia Capital, Arnold Schwarzenegger, Shaquille O'Neal, Henry Kissinger, Google and Tiger Woods.
Applied Scientist
2019 - 2022
Co-developed and launched Amazon Lookout for Vision (https://aws.amazon.com/lookout-for-vision/), a new AWS service for identifying defects and anomalies in images of manufactured products. As part of this: - Developed nearest-neighbor based semi-supervised and supervised anomaly detection models - Contributed to active learning system for customer label set expansion - Built domain-specific object detector based on Detectron2 & demoed to manufacturing partners - Developed a low-shot benchmark for model evaluations
Amazon Web Services provides information technology infrastructure services to businesses in the form of web services.
Research Intern
2018 - 2019
Large-scale video analytics. Designed system for efficiently executing video queries on large camera networks. First author on papers at HotMobile 2019 and SEC 2020 (Best Paper Award).
Microsoft Advertising provides digital advertising solutions.
Graduate Student Researcher
2017 - 2019
Accelerating deep neural net inference on video via feature warping, optical flow, H.264 block motion vectors, and feature fusion. First author on papers at CVPR '19 (oral) and ECCV '18 workshop.
UC Berkeley RISELab develops technologies that enable applications to make low-latency decisions on live data with strong security.
Software Engineering Intern
2016 - 2016
Re-architected full text search feature backend (Apache Solr instance) as a multi-tenant containerized microservice in Kubernetes. Shipped product update to 1000s of live users. Received full-time offer.
Databricks is an AI cloud data platform that interacts with corporate information stored in the public cloud.
Raised $3,997,000,000.00 from Firestreak Ventures and Bossa Invest.
Founder and CEO
2014 - 2015
Developed video-link messaging app for iOS and Android. Reached over 2000 users in 80+ countries. Roles: built iOS app, backend, web presence. Hired & managed Android engineer.