About

I joined Together AI in August 2023 as its first product manager. In those early months I shipped the first self-serve inference products and answered developers on X myself, usually with a docs link and a fix. Now I lead product growth. Before Together I was a growth PM at Asana. Before that I worked on wildfire software at Peregrine, and before that I started a natural-soda company in Costa Rica. I studied Symbolic Systems at Stanford. On weekends I build small things with AI, and this site is where I keep them.

Career

1. 2023–now

   Founding PM → Staff PM → Product Growth Lead · Together AI

   First PM. Helped launch serverless and dedicated inference, fine-tuned (LoRA) and bring-your-own-model hosting, and an OpenAI-compatible API so developers could switch to open models like Mixtral by changing a base URL and a model name. Now I lead product growth.

   Learned: Early on the job was whatever was on fire that week: a support ticket, a docs gap, a launch with a hard date. I got comfortable shipping at 70% confidence as long as there was a clear way back.

2. 2023–now

   Co-founder · Babel

   Dinners, talks and salons for AI founders, engineers and researchers in SF. Started with Abhay Venkatesh and Hari Iyer in October 2023.

   Learned: A community doesn't run itself. When nobody hosts, it goes quiet. Our best nights were a small table with the right few people.

3. 2021–2023

   Growth Product Manager · Asana

   Ran A/B tests on onboarding, team creation and sharing. Shipped new user profiles, my favorite growth win from that time.

   Learned: Most growth wins were small and a bit boring: a profile, a share modal, a better first team. I learned to trust the test over my taste.

4. 2020–2021

   Deployment Strategist · Peregrine Technologies

   Owned the roadmap for NG SCOUT, the platform Cal OES uses during wildfires, and trained the people using its two largest deployments.

   Learned: Fire teams didn't care how the platform worked until they saw what it did for them on a fire day. So I started every training with that.

5. 2016–2019

   Founder · Fresquiticos

   Natural sodas with Costa Rican flavors, sold in 70+ stores, including two supermarket chains. I gave a TEDx talk about it.

   Learned: Selling was mostly showing up, store by store, until the chains said yes. Closing it to focus on Stanford was the hardest call I'd made.

School

B.S. with honors in Symbolic Systems, minor in Data Science, Stanford (2020). Thesis on cultural variance in language intention with Christopher Potts. Mayfield Fellow, and president of Stanford Women in Business.

My system

1. 01I run a set of small bots instead of one assistant. Code, life admin, learning, health and fun each get their own, with one job each.
2. 02Memory lives in Notion. Every bot reads from and writes to the same Digital Brain, so I never explain myself twice.
3. 03Bots draft, I send. Nothing goes out in my name, gets merged or gets bought without my yes.
4. 04One small improvement a day. Each weekday morning a bot picks one fix in my code and asks before it pushes.