Towards systematic breakthroughs via personal AI tools for scientists and investors (an experiment in public VC).Community: www.context.fundProducts: verificationgpt.org

Invest in great companies, especially those in women's health, D2C health, genomics, perception and deep RL.

First Cap is a wealth management group of nice graduates of top PhD programs (Stanford, MIT, etc.), successful entrepreneurs and financial professionals building scalable, diversified investment strategies even our parents can invest in. We're mostly from quantitative founder backgrounds and use growth-oriented long-term strategies. In aggregate, we have over 70 members who collectively control >$100MM in assets and have been growing for over 3 years through several macro cycles.Personal Investments:- Growth: SpaceX, Impossible Foods- Angel: Orca.so (tokens), inito.com, Samos Insurance, Alternativ, Engage Bio, Switch Bioworks, VitaDAO (tokens), vCreate (generally Stanford + YC friends).

How do we make scientists and engineers 100x more productive to solve problems that matter?- Co-founded CodeCompose- Research on contrastive value theory and self-learning for LLMs

- Unlocking human potential via personalized language learning modules for job seekers, powered by AI. 2 years of hard ML x early product and engineering.- Launched in Indonesia.- Press: https://www.youtube.com/watch?v=CklGdMyLIlM&t=2502

Arthena is the first company to offer pricing guarantees for art pieces. These pieces sell for between$1MM - $10MM.- Improved accuracy of core art pricing task ~30% relative in a month. New architecture requires fewer features than prior model and can be trained ~60x faster. Prior model had been in development for almost a year (moderately strong baseline).- Model has been integrated and is being used by the team. Model improves expected profit margin forthe investment product by X2% absolute.- Architected a multi-year deep learning roadmap describing the path to strongly exceeding humanperformance.- Was structured as contract work - returned to Google specifically for the language learning project(above).

Helped start company supporting women through the menopause transition- Grew team from two to five experts (5-20 years’ experience in their respective fields).- Found a well-loved product - the single-most upvoted post on the menopause subreddit by ~2x:https://www.reddit.com/r/Menopause/comments/9z7ifu/menopause_forum_moderated_by_friendly_obgyns/- Got the great lead designer on board back when it was just two people and the seed of an ideavia 2 months of building MVPs and talking to customers.Product launched and grew (>20 on staff at peak):- https://www.bizjournals.com/sanjose/news/2020/11/11/the-cusp-telemedicine-menopause-womens-health.html- https://www.economist.com/international/2019/12/12/millions-of-women-are-missing-out-on-hormone-replacement-therapy

Making healthcare more efficient. Launched menopause startup (above).

Built world-class voice search for e-commerce companies. Imagine using your voice and a cellphone to search and purchase products - that's the experience that we made possible. Competed against the best in the world, made customers happy but didn't have a path to winning the market (backed by YC, Pear and Zhenfund). Wrote a beautiful technical stack that I still miss.We've released an open-source keyword spotter and 700MB dataset from our 2016 NIPS paper: "An End-to-End Architecture for Keyword Spotting and Voice Activity Detection"https://github.com/mindorii/kws

Tracked noisy pulse rates on a watch deployed to 10,000 volunteers in the Google Baseline Study. Primary machine learning scientist and wrote full-stack and infrastructure features as well (compression algos, data pipelines, UI, etc.). Published 2 patents on machine learning and EE techniques for the work. Project spun off into what has become Verily.

CS 229 (twice). CS 231A. CS 110. CS 155.Gave lectures and held office hours for machine learning, computer vision, computer security and computer architecture and design courses.

Kept the stack healthy. Shipped a Scala web service for deployment and version control used throughout the company.Allowed non-developers to query millions of user profiles in seconds rather than hours. Shipped another new service.Assisted Chinese/English negotiations for the first Netease.com partnership, consulted on design of first i18n framework.

Increased translation coverage by 12% for O(10^8) strings on the Facebook platform. Received intern award for the work.