Shrey Gupta

I'm an engineer based in San Francisco building fun & thoughtful AI products.

Olive Campsite NASA Fundsy Irvine Lights Capital One

Olive

For all of 2025 and much of 2026, I worked on Olive as co-founder & CTO. It took on a few shapes over time, but the core product we shipped & sold was AI-native internal tools.

My close friend Bardia and I decided to start a company after a great experience working on Fundsy together. We were both in school and getting more interested in our experiments, so we decided to go full time on it. During the YC W25 batch, we launched as "Lovable for internal tools."

It was exciting to build a cloud platform where you could use AI to build tools connected to your database. A company's database is one of its most prized assets, so we cared a lot about the safety of our code & infrastructure, but also what controls the AI agents we built had. When we started, there was no Claude Code and the best model for coding was Claude Sonnet 3.5.

We loved working with customers with around 20 employees. Ones we could talk with on Monday, prototype something in a few days, ship by Wednesday, get feedback & iterate & ship again by Friday. We slowly realized this type of product worked best in enterprise environments. There was a distinct moment in Orlando at a business & data leaders conference where we realized six-month sales cycles and company politics would hurt that iteration speed.

We decided to pivot a space with unmatched iteration speed: consumer apps. It was exhilarating building & publishing apps that our friends could use. It made us closer to our friends in a way AI B2B SaaS could never. Over five months, we shipped two apps.

The first was an app to find great ideas & inspiration from the web. We wanted to do away with the brainrot consuming all of our social media. There was a time when the internet made you feel inspired, from funny blogs to detailed technical deep dives. We wanted to put those behind a recommender system and put those on your phone as a morning routine.

Our second app was called Treehouse: FaceTime with games. I started using Zoom to call my girlfriend because it was better for sharing music. Bardia and his friends would play games on FaceTime and it was a weirdly difficult experience. We were big Houseparty users back in the day so this was a very nostalgic project of ours.

Irvine Lights

I started Irvine Lights as a freshman in high school. It began as Kidz Club, a science camp where four friends and I ran STEAM events at local community centers.

Before sophomore year, I filed the 1023-EZ myself and turned it into a nonprofit. I posted one Instagram story looking for people to start chapters at their schools and ended up with 25 student organizers across Irvine. We ran camps, workshops, and larger community events for hundreds of kids.

During COVID, most of that stopped working. I started using Irvine Lights as a fiscal sponsor instead, helping other students start community projects under our nonprofit. We handled their donations, budgets, reimbursements, and finances, which eventually became the problem that led us to build Fundsy.

Fundsy

Fundsy was a platform to start & manage community projects.

I founded it at age 16 with my best friends Bardia & Eric. None of us had built an app before. We came together as a team for the Congressional App Challenge.

The idea came directly from running Irvine Lights. I was fiscally sponsoring community projects for a few friends. Donations would come in through PayPal, and I would manually reconcile every dollar, manage budgets, and process reimbursements.

We built Fundsy to handle all of that in one place. Projects could accept donations, track their balance and budget, submit reimbursements, and issue cards to team members for spending. We built the card product on Stripe Issuing, which was still in beta at the time.

Each project had its own finances, but everything still rolled up to the fiscal sponsor. That meant we had to keep track of which money belonged to which project, who was allowed to spend it, and make sure every transaction was accounted for.

We won the Congressional App Challenge and ran Fundsy for three years. A lot of customer support happened while we were walking our dogs.

NASA

At NASA, I worked on the SmallSat Data Acquisition team as a Software Engineering Intern. We handled a huge stream of satellite imagery, nearly 200 petabytes per year, which was distributed to licensees including government departments and university researchers.

I worked on a feature that let licensees run analysis and computation on NASA's servers before transferring the data out. Instead of downloading massive amounts of imagery and processing it themselves, they could filter for only the data they actually needed. Because compute was relatively cheap and far fewer images had to leave NASA's network, this would save millions of dollars in egress costs.

For example, a researcher could identify which bodies of water had grown or changed color, then download only the relevant imagery for further analysis.

Campsite

Campsite was a five-person team and some of the most talented people I've ever worked with. It was an async team communication platform that focused on deep work and independent thought. Surprisingly, this had a significant, positive second order effect on speed. I modeled a lot of how we worked at Olive after Campsite.

As a software engineering intern, my big project was to build out the "Follow Ups" functionality & integrate it across the app. I came in with little TypeScript & no Ruby on Rails experience, and came out with a deep understanding of React styling, portals, Rails & DB modeling. It was the first place I learned how production software actually works (migrations, backfills, N+1 queries, job queues, SSO, etc).

Capital One

At Capital One, my team worked on a customer retention module for call centers. Call center agents could take cute life moments from things people say on calls, like a customer mentioning their son graduating, and send them a small surprise gift.

This program was called Surprise & Delight and it had been running for 5 years before we built this, but everything used Google Forms and operations staff to do the purchasing manually. We directly integrated it in their software and automated it.

It was funny since our product actually had Capital One spending more money.

I lived in Richmond, VA, which had surprisingly great food and concerts.