A tool to automate testing and comparing ChatGPT responses based on variations in prompts and configurations
- Supported the Presideo Creative agency to complete their product recommendation engine.
- Built tool to test performance and recommendations of different use cases, comparing OpenAI models and methods - embeddings, Completions, and Assistants - preserving responses back to prompts and configs for fine-tuning.
- Created front-end visualization table to filter, sort, compare, and annotate results.
- Wrote python scripts to pre-generate embeddings from internal product lists
