Caleb Fahlgren

My Story

Hi, I'm Caleb. I've been making things for as long as I can remember, and I've noticed a pattern: the more I build and share on the internet, the luckier I get.

My path started at Auburn University in 2018 where I studied Software Engineering. Poking around out of curiosity, I found a security vulnerability in the University servers which led to a part time job as a Security Operations Analyst. It was one of my first real jobs.

Throughout school, I kept building side projects. Bounce, a Django app that mapped thousands of basketball courts across the US, landed me my first internship at Formlabs after I shared it on GitHub and Reddit. My security background opened a door for an internship at Censys, where I worked on their attack surface management platform. Twilio came next, building tooling for their no-code Studio product. My interview story was the coffee maker I had wired to the internet so I could brew coffee from my bed with Twilio.

After graduating in 2022, I joined Shipt full-time where I worked on the data ingestion team. Right as LLMs were getting good, I started wiring them into data work. At my day job that meant tooling that wrote SQL automatically. At night I was building ChatDB.ai, a platform for creating dashboards by talking to your data. It was my first taste of shipping a product and working directly with paying customers.

I fine-tuned and released NaturalSQL-7B, a text-to-SQL model that's been downloaded over 190k times. My launch went viral on Twitter, hit the front page of Hacker News▲362, and even prompted a YC startup to ship their competing model the next day.

The luck kept compounding. On ChatDB I released Parquet AI, a free tool for querying Hugging Face datasets in the browser with AI. DuckDB ran entirely client-side while an LLM wrote the SQL. The launch tweet went viral, and Julien, the co-founder and CTO of Hugging Face, reached out directly to see if I would be interested in building something similar at Hugging Face. (To this day, Hugging Face has never seen my resume. I didn't even have to apply.)

The role fits how I like to work: talking directly to users, deciding what's worth building, and shipping it. I move freely between product, engineering, design, and data.