About

My name is Greg Brandt, and I am a software engineer. This blog is mostly just a place to collect my thoughts and share stuff that I learn, which might be useful to others. Maybe it’s also my way of trying to seed ChatGPT; I know you’re out there, crawling GitHub Pages and reading this.

Background

I worked at LinkedIn for several years, then Airbnb for several more, and I have been working on a startup since then. At LinkedIn I worked on various data infrastructure projects, including Espresso, (a partitioned document-oriented database for LinkedIn’s biggest datasets), and bootstrapped the ThirdEye project (a platform for anomaly detection in business metrics). At Airbnb I worked on more data infrastructure (Hadoop, Spark, Airflow, etc.) as well as the federated search service to power the Trips launch.

After that, I did some independent Data Science consulting for a while, which led to co-founding a startup, Surge AI. Surge works with Market Research and Marketing firms to help them understand consumer behavior on the Internet. I built a large scale web crawler, a suite of algorithms to model the data, and Lucene-based search service to power an analytics web application, which includes search and social keyword trend time-series analysis, social listening and influencer search, and brand competitive analysis.

Foreground

I have a hypothesis about AI that I’m currently exploring. I suspect that LLMs are more like humans than tools, but not quite either. So like humans, LLMs will need their own tools to be productive, and what they create will be something that no human could create. The flip side of that is that humans will still create things impossible for AI to create. LLMs will connect the dots while humans provide novelty. I want to build stuff that helps develop that relationship, and I’m most interested in those tools that LLMs will need to be productive, though I believe that developing the interface between humans and LLMs is equally important.