skip to content

Nicolay
Christopher Gerold

About

Hi, I’m Nicolay. I’m an AI engineer specializing in building practical AI solutions with LLMs.

A cartoon cactus looking at the 'Astro.build' logo

I've been working with LLMs since 2019. I've seen what works and what doesn't.

I started in 2019 finetuning encoder-decoder models for clustering and classification and quickly discovered the GPT-family First just for fun and learning, but after the GPT-3 release I really recognized the potential of the models and went all in and started to freelance in LLMs. I started out with a lot of use-cases around information extraction in healthcare and insurances and then went on to build some early tries at agents for financial report generation.

Now, I mostly build complex workflows or agents taking over specific tasks through a combination of search, traditional machine learning and foundation models.

Some of the stuff I've built:

  • LLM-powered search for a large insurance company. 15% increase in nDCG and sped up time to contract by 5%.
  • Agent for financial report analysis for a Swiss bank to improve decision making support.
  • LLM-powered data extraction for a healthcare startup tackling electronic health records. Improved accuracy by 12% and reduced cost by 24%.
  • Agent for M&A due diligence replacing 100+ hours of manual work.
  • Automated customer support classification system for a telecom provider. Reduced response time by 67% and improved first-contact resolution rate from 76% to 91%.
  • Semantic search engine for legal documents, processing 2M+ contracts. Reduced discovery time by 67% and saved $200K in annual review costs.

I have my own podcast where I interview the best builders in the AI space. How AI Is Built

On occasion, I also jump on a podcast to talk about AI, mostly about LLMs and AI agents: