Ecosystem & Stack: gpu

Streamlines complex deep learning engineering, enabling scalable AI model training and finetuning across diverse hardware with minimal code changes.

A comprehensive platform enabling users to create personalized AI twins by fine-tuning Large Language Models with their unique chat history, capturing individual style and bringing digital selves to life.

A GPU-accelerated, scalable toolkit for multimodal data preprocessing and curation, designed to train better AI models faster.

A high-performance, scalable JAX-based open-source library for training large language models on Google Cloud TPUs and GPUs.

Optimizes large language model inference to run 70B models on a single 4GB GPU without quantization, enabling efficient deployment on resource-constrained hardware.

An asynchronous reinforcement learning framework enabling personalized AI agent training through natural language conversations and scalable real-world deployments.

Docta is an advanced data-centric AI platform that detects and rectifies issues in various data types to improve model performance.

This project implements Google's Dreambooth technique on Stable Diffusion, enabling users to fine-tune a text-to-image model with a few custom examples for personalized image generation.

Parlor is an on-device, real-time multimodal AI that enables natural voice and vision conversations, running entirely on your local machine.

A high-performance, no-abstraction toolkit providing recipes for pretraining, finetuning, and deploying over 20 large language models at scale.