Hey PH'ers! I was the Search/NLP specialist at MongoDB for 3 years, and learned A TON about how to architect robust search architectures. I built Mixpeek as the outcome of these lessons, which uses state of the art Large Language Models (LLMs), Data Structures (HNSW), and Architectures (fully Serverless) to achieve indexing, searching and relevance better than what's off the shelf. We're venture-backed and actively looking for design partners. If you use S3 and have a large variability in filetypes please reach out. The API has a python library and is currently in open beta. If you plan on adding it to your software, please send me an email ethan (at) mixpeek (dot) com so I can ensure we provision hardware appropriately.