I built an LLM pipeline to measure political sentiment across large text corpora, part of the Digital Markets Initiative's work on the economic and political footprint of foundation models. It runs on UF's HiPerGator supercomputer, generates sentence embeddings, and uses cosine similarity to cluster and classify sentiment at a scale you couldn't do by hand. I was less interested in the noise around LLMs than in what they actually let you measure once you point them at real text.
This is where models leave the notebook and have to survive contact with real users. I build recommendation and ranking systems that actually run in production: trained, backtested behind a quality gate, and served over HTTP. Plus the LLM pipelines around them. Most of the work isn't the model; it's making it reliable, reproducible, and fast enough that a product can lean on it. I don't call myself an 'AI engineer', I just want to understand these things well enough to build with them.
Browse the full collection →