AI Infrastructure

QKeras provides quantization tools for deep learning with TensorFlow Keras.

LLM Gateway routes, manages, and analyzes requests across more than 40 language-model providers through one API.

Adaptive Reliability Layer decides when to retrain a model based on drift signals and delayed labels.

ColossalAI provides tools intended to make large AI models cheaper, faster, and more accessible.

GeneticSharp provides C# tools for developing applications that use genetic algorithms.

Latitude traces AI agent behavior in production, finds failures, provides alerts, and can dispatch coding agents to propose fixes.

Hyperopt supports distributed asynchronous hyperparameter optimization in Python.

The linfa project provides a machine learning framework written in Rust.

Kitaru provides agent traces that teams can run and inspect during application workflows.

autobatcher batches asynchronous OpenAI-compatible API requests through a drop-in replacement library.

Hopsworks provides a feature store and MLOps platform for building and operating production machine learning systems.

Machine learning operations and large language model operations are the focus of Starwhale’s platform.

AI/ML API provides access to models for text, images, video, audio, code, and embeddings through one API.

A JavaScript Native PyTorch-aligned Machine Learning Framework, built from scratch on WebGPU.

Vercel’s AI SDK is a TypeScript toolkit for building AI applications and agents with React, Next.js, Vue, Svelte, and Node.js.

Metorial connects employees, agents, and AI clients to company tools and skills with access controls and tracing.

Tasklet connects business tools and delegates operational work to AI agents through an agent-based business platform.

Chutes provides serverless compute for deploying, running, and scaling AI models through its platform and API.

Manifest connects AI agents and tools to multiple model providers through an LLM gateway.

List of awesome compiler projects and papers for tensor computation and deep learning.

Kimi K2 handles complex language tasks, reasoning, and problem-solving through an AI model developed by Moonshot AI.

File Parser optimised for LLM Ingestion with no loss. Parse PDFs, Docx, PPTx in a format that is ideal for LLMs.

Run sandboxes, task queues, and custom model inference with ultrafast boot times, instant autoscaling, and a developer experience that just works.

Activeloop provides infrastructure for continual learning and develops the Deeplake GPU database for agents.