Endeavours

Operating

AI × Computational Science

I write most of my research scripts using LLM chatbots. They're surprisingly good at abstract reasoning, but struggle with context—I spend as much time reteaching the bot about project details as I would writing the script myself. The deeper into a project I get, the more context there is to manage, making off-the-shelf chatbots impractical for serious scientific work.

Can we build an agentic LLM interface that solves this by intelligently routing the right context to the right tasks? How might "programming" context like code allow us to automate more of scientific problem-solving, freeing researchers to focus on asking better questions rather than technical implementation? Instead of trying to replace scientists, can we empower them to do more with less, especially given the current funding environment? These are some of the questions I'm exploring at South Park Commons, with the goal of accelerating scientific progress by turning every grad student and postdoc into a PI managing their own team of LLM research agents.

Networked Identity (Sora ID → CLEAR)

Today, we predominantly verify our identities online using things we know (passwords, SSNs, magic codes, security questions) or things we have (devices, ID cards). But if a fraudster knows your secrets or spoofs your credentials, they have your identity, and if your identity is compromised at one place, it's compromised everywhere. This fundamental gap in Internet infrastructure leads to 3 million Americans getting their identities stolen every year and hundreds of billions of dollars stolen or laundered, much of which goes towards funding terrorists and rogue states.

Prompted by rising fraud during COVID and growing adoption of technologies like Apple's FaceID, I cofounded Sora with Stanford classmates to solve this problem by building biometric identity credentials for every user (with consent, of course!). Sora's platform verified user identities for fintechs and banks, then issued strong biometric credentials that could be reused at any other Sora customer. We raised $6M from investors including Greylock, Abstract, and General Catalyst, scaled the business to seven figures of revenue and hundreds of thousands of users, then sold the company to CLEAR (NYSE:YOU) in 2023. Now, Sora's technology powers CLEAR's identity network across sectors, serving over 30M Americans and household names such as LinkedIn, Home Depot, Avis, Uber, T-Mobile, and Docusign.

If you're even more interested in the Sora/CLEAR story and have half an hour to kill, here's a podcast on which I talk through the journey.

Investing

I am an active angel investor with a focus on early-stage startups. I especially enjoy advising young and/or first time founders, where I aim to be the steady voice I wish I had during my first rodeo. Beyond my own check, I've helped my portcos find and close candidates and customers, raise additional financing, and think through strategic decisions. Select investments include:

• Wispr Flow: do-it-all AI voice assistant. Invested at Seed alongside NEA and 8VC.
• Slingshot AI: foundation models for mental health. Invested at Series A alongside a16z.
• Abacus AI: copilots for every accountant. Invested at Seed alongside Menlo.
• Crippling Hot Sauce: hot sauce for every underdog, led by Drew Davis, a 19 year old entrepreneur with cerebral palsy. Invested at Seed.

I also operate a small fund with a friend, where we translate what we're seeing in the early stage market into growth and public bets.

Advising

I am an advisor to Oncare AI, which uses AI to optimize operations for leading cancer centers. I also occasionally join advisory boards through StartX, which Sora was lucky to be part of in Winter '21.

On a more ad-hoc basis, I've guided several friends through their Seed raises, helping put rounds together with Abstract, Kleiner Perkins, Craft, and several other leading venture firms. Feel free to reach out if you're raising.