Own:- Machine Learning team: building AI-assisted labeling features and processes to improve data quality and throughput- Perception team: building multi-modal editors (language, CV, audio, video etc) to label a variety of use cases across gen AI

Own all Computer Vision offerings

Over a period of 12 weeks, underwent a rigorous exploration of VC deal sourcing, due diligence, alternative investments, mechanics of preferred stock and convertible notes, term sheet negotiation, and how to think about dilution in the context of a VC financing.

Joined with a mandate to integrate Computer Vision technologies into Palantir Gotham, via acquisition of Synapse Technology.

The company was acquired by Palantir Technologies in 2020.As the first CV and ML engineer at Synapse, my primary responsibility is to research, design and build a proprietary artificial intelligence system, bringing together my backgrounds in computer vision, machine learning and software engineering. Over the last 2 years, I have been responsible for:- Working with the CTO to design and implement the company's AI strategy, and lay the foundations for the Perception Team (as we call it internally)- Researching new research papers to take advantage of the latest advances in the field- Training and deploying deep learning models for object recognition- Building an end-to-end computer vision pipeline, including data collection, data labeling, data understanding, model research, model training, evaluation and on-the-edge deployment at real paying customer locations around the world such as the Osaka airport and Singapore- Recruiting new computer vision engineers- Managing 2 interns into progressing in their machine learning careers- Working with the core software team to design the system that enables our neural networks in the field- Conducting technical and sales trainings for Synapse's distributorsSynapse Technology Corporation provides algorithmic analysis of security X-ray images to determine if they contain a threat, such as a firearm or explosive. The company deploys their deep learning based algorithms on premises at global airports, schools, federal buildings, and provides a user interface for screeners to communicate with the algorithms.

Doing research and taking classes in Artificial Intelligence and Human-Computer Interaction.Strongly interested in the intersection of the fields, such as human-in-the-loop systems and interactive machine learning to make ML more interpretable and accessible outside the field.Recent paper on reinforcement learning applied to build an intelligent math tutor accepted for poster presentation at AAAI Conference on Human Computation and Crowdsourcing (HCOMP) 2017. Read an earlier draft at: http://stanford.edu/~jayhp9/phipaper.pdf.

Intelligent programming assistant which connects to your favorite editor. Kite infers types in your code and recommends documentation, code examples and error fixes to help you code more efficiently. Unlike an IDE, Kite is backed by programming knowledge from all over the web.

Project: Expert Crowdsourcing with Flash TeamsWorking with Assistant Professor Michael Bernstein of the CS DepartmentResponsibilities (mainly coding):Built real-time capabilities (with Javascript) into a single-page web interface that allows skilled teams to collaborate seamlessly on projects.Currently building a Javascript API to enforce dependencies between each team member's tasks.

Topics taught: techniques for controlling complexity; strong modularity using client-server design, virtual memory, and threads; networks; atomicity and coordination of parallel activities; security, and encryption; and performance optimizations.Responsibilities: Grading, holding office hours.

Topics taught: Concepts and techniques used in constructing interactive web applications. Browser-side web facilities such as HTML, cascading stylesheets, javascript, and the document object model. Server-side technologies such as sessions, templates, relational databases, and object-relational mapping. Issues in web security and application scalability. New models of web application deployment.Responsibilities: Teaching section(s), grading, holding office hours.

Team: SearchAccomplishment: Increased number of facts contributed to Knowledge Graph (KG) via GRASP by 30%, by improving the classifier component. The improvements ranged from large design changes to tuning classifier parameters to obscure bug fixes. GRASP is a pipeline that determines if extractions from the web are facts and adds only the high-precision ones (>99%) to KG.

Wrote a Chrome extension (in JS) from scratch to improve users' listening experience by tracking their searches on the web and recommending playlists, and providing keyboard shortcuts to control the music playback from any tab on the browser.Wrote Hive/MapReduce scripts to retrieve specifc data about users for targeted marketing purposes. For example, all users who have played at least one song of a particular language in the past 6 months.

Helped organize and lead the first two TEDxStanford events in 2012 and 2013.

Wrote an internal search tool from scratch in Ruby on Rails to query more than 300K users based on 16 different parameters across 3 different Mongo databases.Wrote a feature that allows Piazza users to unsubscribe from promotional emails.

Optimized an in-house developer framework, by rewriting inefficient code.Created a website to showcase the framework.

Right out of high school. Built a virtual gaming world from scratch, which raised more than $400K in seed funding.