My time at Pulley

I never saw myself working anywhere else except the startup. Everyone around me was getting jobs from big tech and quant firms. What I wanted was to work at a small startup; I wanted to be alongside people who are risking it all and building something from scratch. Being surrounded by these kinds of people, doing something that has a very low probability of success, fills me with immense energy and makes me the happiest person ever.

So, as everyone around me at MIT was chilling with their offers from the corporate world, I was spending my nights going through the YC Startup Directory and cold-DMing every founder of the startup that seemed interesting. I just love cold messaging people. This is my favorite thing in the world.

400+ messages resulted in 10 calls, which resulted in 2 offers.

As I was interviewing with companies (spring of 2023), GPT-4 had just come out as an API, and Mark Erdmann (co-founder at Pulley) said that my job for the summer would be working together with him and trying to build many prototype tools with GPT-4 for Pulley. That sounded like something I would do even if I wasn't paid, so I immediatelly accepted the offer from Pulley.

Due to the NDA, I cannot go into details of what we worked on. But I can say that there were two categories of tools:

First, internal tools. These ranged from RAG tools to speed up productivity to LLM agents that can handle parts of the workflow previously done by humans. The biggest hit was the company Slack bot, powered by RAG on the Slack history and company's documents. It was queried 30-50 times a day by employees, and everyone loved it.

Second, user-facing tools. I developed many prototypes on how LLMs can be used in the company's existing user-facing tools. Some of these prototypes were converted to actual user-facing tools that bring revenue to the company.

I'm really grateful to everyone at Pulley, specifically Mark, Yin, and Grant for the support, freedom, and mentorship. I would strongly recommend anyone to work there.

My strategy of just taking the YC Startup Directory and DMing the founders clearly worked. It gave me the ability to experiment and work hard with LLMs in real-time as they were taking over the world. I recommend this strategy to everyone here at MIT. So far, 5 of my friends got a startup job the same way.