Project Tapestry aims to give every nation and participant frontier AI they can call their own — uniting a global consortium to train a shared frontier model from which partners build and own sover…

OpenTela is a decentralized compute fabric for running machine learning applications.

A Rust framework for correct and performant distributed systems

AI model training on heterogeneous, geo-distributed resources

Redset is a dataset containing three months worth of user query metadata that ran on a selected sample of instances in the Amazon Redshift fleet. We provide query metadata for 200 provisioned and s…

CUDA benchmarks for measuring GPU utilization and interference

Dirigent: Lightweight Serverless Orchestration

Microsoft Azure Traces

A lightweight, user-friendly data-plane for LLM training.

Rust-like error handling in Go

Lightweight Helm IDE written in Rust with the Tauri framework... it also uses HTMX btw

Load generator and trace sampler for serverless computing

Compression for Foundation Models

An interference-aware scheduler for fine-grained GPU sharing

Utilities for Training Very Large Models

Modyn is a research-platform for training ML models on growing datasets.

Elastic computing platform

ML Input Data Processing as a Service. This repository contains the source code for Cachew (built on top of TensorFlow).

ML Input Data Processing as a Service

Resources for Data Centric AI

DuckDB is an analytical in-process SQL database management system

vHive: Open-source framework for serverless experimentation

SeBS: serverless benchmarking suite for automatic performance analysis of FaaS platforms.

Reference implementations of MLPerf® training benchmarks

Machine Learning Toolkit for Kubernetes

Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

Pocket metadata and data plane (based on Apache Crail)

Elastic ephemeral storage