bioinformatics/computational biology. Specifically, multiple sequence alignment and HMMs for protein recognition. The author was previously a physicist (PhD) who left physics due to the low number of jobs in the 70s-80s, founded a database company, sold it to Intel, and then visited Berkeley and saw a cool talk and volunteered.

There were already a few codes in the area and plenty of papers, and he was mathematically inclined, so it didn't take long for him to become an expert. Once he was an expert, he pointed out major problems in existing codes (both functional and performance).

This is an area where you're working with fairly straightforward data and math (linear strings from a chosen alphabet, probabilistic model is well-established). You don't need to understand the underlying biology in detail to contribute.