Collective Knowledge (CK), Collective Mind (CM/CMX) and MLPerf automations: community-driven projects to learn how to run AI, ML, and other emerging workloads more efficiently and cost-effectively across diverse models, datasets, software, and hardware using MLPerf methodology and benchmarks.

MLCube® is a project that reduces friction for machine learning by ensuring that models are easily portable and reproducible.

Pythonic binding to the Apple Neural Engine

Legacy CM repository with a collection of portable, reusable and cross-platform CM automations for MLOps and MLPerf to simplify the process of building, benchmarking and optimizing AI systems across diverse models, data sets, software and hardware

MLPerf (tm) Tiny Deep Learning Benchmarks for STM32 devices

This repository contains automation scripts designed to run MLPerf Inference benchmarks. Originally developed for the Collective Mind (CM) automation framework, these scripts have been adapted to leverage the MLC automation framework, maintained by the MLCommons Benchmark Infrastructure Working Group.

Benchmark OpenAI-compatible AI endpoints and AI Accelerators in a reproducible structured way

MLCFlow: Simplifying MLPerf Automations

nvProbe — Open-source NVIDIA GPU benchmark suite for CUDA workload automation, Slurm HPC cluster profiling, and MLPerf reporting

CIFAR10 training repo for MLPerf Tiny Benchmark v0.7

CM interface and automation recipes to analyze MLPerf Inference, Tiny and Training results. The goal is to make it easier for the community to visualize, compare and reproduce MLPerf results and add derived metrics such as Performance/Watt or Performance/$

Converting models used by MLPerf Mobile working group to Core ML format

A benchmark suite to used to compare the performance of various models that are optimized by Adlik.

Popperized MLPerf benchmark workflows

Development version of CodeReefied portable CK workflows for image classification and object detection. Stable "live" versions are available at CodeReef portal:

Edge AI deployment knowledge graph — boards, accelerators, ONNX operator kernels and quantized models in one graph. Real ONNX + ONNX Runtime + MLPerf Tiny data plus a generated fleet. Built on Samyama Graph.

Tekton Pipelines to run MLPerf benchmarks on OpenShift

Performance benchmark evidence repository — activates at first-silicon bring-up (Sentinel-1 Q4 2026).

Transparent benchmark cards, JSON artifacts, and visual tools for photonic AI accelerator energy/noise claims.

These are automated test submissions for validating the MLPerf inference workflows