Ruby startup demo – Wyatt Ades
summary
An edtech LMS built on Rails runs AI-assisted grading for 5-minute student video submissions using audio transcripts, rubrics and LLM tool calls that produce timestamps for rubric items. Baseline brute-force LLM grading shows about 18% average error and a worst-case rubric item at ~30% error, so the system can automate 80% of grading while routing the remaining 20% to human instructors; goal is 10,000 graduates this year.
AI-written summary · about
