CHAI: AI Platform

CHAI is a generative AI startup located in Palo Alto, building a platform where anyone is free to submit a Large Language Model (LLM). The company leverages human feedback from millions of users to combine individual LLMs into a meta-model that vastly exceeds the capabilities of any single model. CHAI's research focuses on applying proven techniques such as RLHF, SFT, Prompt Engineering, Rejection Sampling, and LLM routing, iterating across various algorithms like DPO, LoRA, training reward models, and embeddings to build recommender systems, all aimed at improving AI for increased monetization and engagement. They also tackle engineering challenges related to scaling LLMs, experimenting with techniques like quantization, custom CUDA kernels, Flash Attention, and KV-Caching, and building custom implementations for Kubernetes clusters and inference engines.