I’m Charlene Chambliss, former machine learning engineer and current software engineer.
I used to work on large language models (LLMs), training them to do my bidding (and by “do my bidding”, I mean “read a lot of words.”).
Admittedly, I got a bit bored of the cycle of training a classifier, or an entity detection model, or a question-answering model, or what-have-you. That code looks pretty similar to the first time you wrote it every subsequent time you write it, and after a while I felt like I wasn’t growing.
Once I joined a product team, though, I got hooked on the challenges of making software and machine learning features work together. I had a great time shipping various ML-powered features with my team at Primer, and greatly enjoyed the whole design and implementation process from start to finish. It turns out that even with LLM magic, good products still need a healthy dose of UX.
How I got into data science, then ML, then software (from a thoroughly nontechnical background) is a long story. Thankfully, a wonderful former colleague wrote a great interview piece where I describe my journey and my reasoning for each step I took along the way, or you can read an excerpted version on this site.
After Primer, I joined Aquarium Learning, where for the first year or so I helped build software to help ML teams curate their datasets and train better models. We recently released a new product, Tidepool, that helps teams with LLM-based products in production figure out how they can best improve their products for users.
This site is built with Astro. As a framework that emphasizes content-friendliness as a key goal, Astro has been a great fit for my small blog property here. The content collection is so easy to use and saved me a lot of time trying to figure out all the content management and Markdown presentation details.
Besides Astro, the site is built with TypeScript and Open Props to keep the CSS and theming simple.