• Listen Labs has built an AI interviewer that can ask follow-up questions, like a conversation, but at the scale and ease of use of a survey. • Listen Labs clusters free-form conversation into a structured format. • We use embeddings, fine-tuned large language models, and more to solve these problems. • We're fine-tuning and prompting in novel ways to get the right output from LLMs. • We’ve built the fastest speech-to-speech pipeline but there’s a number of challenges we need to address to make it better.