Libraries

Google's Agent Development Kit (ADK) provides a comprehensive framework for building, testing, and deploying AI agents. It includes tooling for agent development with built-in functionality for reasoning, memory management, and multimodal interactions.

Scalable deep learning framework that supports both imperative and symbolic programming, enabling flexible development and efficient deployment across devices from cloud infrastructure to mobile devices.

Microsoft's framework for building multi-agent AI systems where multiple LLM agents collaborate, debate, and solve tasks through conversation.

Deep learning framework developed by Berkeley AI Research, focused on expression, speed, and modularity. Particularly strong for computer vision applications and convolutional network models.

Lightweight, modular, and scalable deep learning framework built on the original Caffe. Designed for production deployment and mobile applications. Now part of PyTorch.

Framework for orchestrating role-playing AI agents. Define agents with roles, goals, and tools, then have them collaborate as a crew to complete complex tasks.

Next generation platform for object detection, segmentation, and other visual recognition tasks. Built on PyTorch, it's designed for flexibility and extensibility.

Stanford's framework for algorithmically optimizing LLM prompts and weights. Replaces manual prompt engineering with programmatic pipelines and automatic optimization.

Sequence modeling toolkit that allows researchers and developers to train custom models for machine translation, text generation, language modeling, and other text generation tasks.

High-level API for Apache MXNet, jointly developed by AWS and Microsoft. Makes deep learning easier with simple, concise API for defining models using a collection of pre-built, optimized neural network components.

Open-source NLP framework for building production RAG pipelines, document search, and question answering systems with any LLM or retriever.

Python library for structured outputs from LLMs using Pydantic models. Handles retries, validation, and streaming for reliable, type-safe LLM responses.

High-performance numerical computing and automatic differentiation library developed by Google Research. Combines Autograd and XLA for efficient ML research on accelerators.

LangChain is a framework for developing applications powered by language models. It enables applications that are context-aware, reason, connect to other data sources, and integrate with agents for complex tasks.

Data framework for LLM applications. Specializes in ingesting, indexing, and querying private data with RAG pipelines and agentic workflows.

Collection of reference MCP server implementations for various services including filesystem, GitHub, Slack, Google Drive, and more.

Commercial-grade distributed deep learning toolkit. CNTK allows users to easily realize and combine popular model types and enables efficient implementation and execution of RNNs, CNNs, and feed-forward DNNs.

Open Neural Network Exchange is an open format for representing machine learning models. ONNX defines a common set of operators and a common file format to enable model interoperability between frameworks.

High-quality implementations of reinforcement learning algorithms. Baselines provides reference implementations to reproduce published results and serves as starting points for developing new approaches.

Toolkit for developing and comparing reinforcement learning algorithms. Gym provides a standardized set of environments for agent benchmarking and is compatible with any numerical computation library.

Library for structured text generation with guaranteed output formats. Enforces JSON schemas, regex patterns, and grammars directly at the token sampling level.

A fast, powerful, flexible and easy to use open source data analysis and manipulation tool built on top of Python.

Hugging Face library for fine-tuning LLMs with methods like LoRA, QLoRA, Prompt Tuning, and Prefix Tuning using minimal GPU memory.

An open source machine learning framework that accelerates the path from research prototyping to production deployment.

Simple and efficient tools for predictive data analysis. Built on NumPy, SciPy, and matplotlib, it's accessible to everybody and reusable in various contexts.

Microsoft's open-source SDK for integrating LLMs into applications. Supports plugins, planners, memory, and multi-agent patterns in Python, C#, and Java.

An end-to-end open source platform for machine learning. Comprehensive, flexible ecosystem of tools, libraries, and community resources.

End-to-end platform for deploying production ML pipelines. TFX standardizes the components needed for ML engineering, including data validation, feature engineering, model analysis, and serving.

Lightweight version of TensorFlow for mobile and embedded devices. Enables on-device machine learning with low latency and small binary size.

Run Hugging Face Transformers directly in the browser using ONNX Runtime Web. No server required — supports text, vision, audio, and multimodal models.

High-throughput and memory-efficient LLM inference engine. Features PagedAttention for 24x higher throughput than HuggingFace Transformers.