Product manager builds Beval, a lightweight web app for quick AI evaluation without complex infrastructure setup

Created Beval to solve the pain of slow, cumbersome LLM evaluations on conversation transcripts and agent traces

Previously blocked by limitations of 'Gemini in Google Sheets' which was too slow and handled evaluation changes poorly

Needed simple yes/no evaluations like 'did the agent answer the question' or 'did it cover 5 required points' without building full infrastructure

Initially learned to call OpenAI API in Python as a workaround, but wanted to create a product-level solution for others with similar needs