Rockwell Windsor Rice is a senior full-stack engineer with 10+ years building production Rails and Next.js applications for SaaS companies and government systems. He's maintained 15+ production applications as solo engineer at Levee Labs, modernized mission-critical APIs serving 100,000+ veterans daily on VA Lighthouse, and taught 200+ students as Udacity Front-End instructor.
In his spare time he can be found playing classical guitar or having fun with his best buddies and .
* You can chat with whichever puppy is present by trying out the AI chatbot I built - chat with Gus or Mitch! Each one has a personality of its own. Read all about how it works here
A walkthrough of puppy-evals, an evaluation harness for the RAG chatbot at windsordevelopmentstudio.io. Covers the three-judge architecture, a golden set of 60 test questions, and the results of two experiments that improved retrieval from 86.7% to 100% and persona consistency scores by 0.66 points.
5 min read
AI models like Claude and GPT aren't finished tools but raw materials developers shape into custom solutions for their specific workflows. By treating large language models as building materials rather than ready-made products, developers can craft purpose-built integrations that enhance their existing practices in Next.js, Ruby on Rails, and JavaScript development.
15 min read
