Link Search Menu Expand Document
AI Alliance GitHub Organization

AI Alliance GitHub Organization

Quick Links

The AI Alliance Projects

Welcome to the GitHub organization for the AI Alliance, a community of technology creators, developers, and adopters collaborating to advance safe, responsible, and effective AI rooted in open innovation. Here, Alliance members collaborate on technical initiatives, AI technology guides, and related projects.

The AI Alliance is focused on accelerating and disseminating open innovation across the AI technology landscape to improve foundational capabilities, safety, security and trust in AI, and to responsibly maximize benefits to people and society everywhere.

See about us for more information about The AI Alliance.

Focus Areas

The Open Agent Lab: Collaborate, Experiment, and Build Domain-specific AI Agents with Open Source

The Open Agent Lab is a collaborative community, open source projects and domain-specific work groups that seek to make AI Agents successful in the real world through fast experimentation and distillation of learning into usable reference architectures and implementations, and build out of new tools to enable development and deployment.

We are especially focused on domain-speciic challenges with work groups in:

Industrial AI: check out SemiKong a foundation model for semiconductor process agents.

Legal: try Bartlebot if you want to work with case law and other legal topics.

Climate and Geospatial: check out projects like GeoBench and TerraTorch.

Chemistry and Materials: take a look at new science foundation models for molecular analysis.

(Coming soon.) Finance, Health, and Time Series domain work groups.

The Open Agent Lab builds with open weight models from our members, including Llama, Granite, AI2, ServiceNow, and more.

Agent Projects

Links Description
AllyCat AI Alliance icon
(Beginner friendly!) Get started with a simple and fun end to end RAG application that scrapes your website so you can ask it questions.
Agent Lab UI AI Alliance icon
AgentLabUI is a web-based interface designed to simplify the creation, management, and deployment of AI agents. It supports integration with Gofannon tools, enabling developers and researchers to rapidly prototype and experiment with sophisticated AI agent architectures.
Gofannon AI Alliance icon
A repository of functions consumable by other agent frameworks.
Proscenium AI Alliance icon
A lightweight, composable library and several demo applications.
Llama Stack Agents AI Alliance icon
A deployable, domain-specific application demonstrating the use of Llama Stack.
MCP Gateway
A Model Context Protocol (MCP) Gateway. Serves as a central management point for tools, resources, and prompts that can be accessed by MCP-compatible LLM applications. Converts REST API endpoints to MCP, composes virtual MCP servers with added security and observability, and converts between protocols (stdio, SSE). (Principal developer: IBM)
The Living Guide to Applying AI AI Alliance icon
Tips from experts on using AI for various applications, including popular design patterns.
AI Application Testing for Developers AI Alliance icon
If you are a software developer, you are accustomed to writing deterministic tests. What do you do when generative models aren't deterministic?
OpenDXA (coming soon!) AI Alliance icon
Domain Expert Agents (DXA) for industrial AI. (Principal developer: Aitomatic)
Docling
Docling simplifies document processing, parsing diverse formats — including advanced PDF understanding — and providing seamless integrations with the gen AI ecosystem. (Principal developer: IBM Research)

Llama Stack

(coming soon to AI Alliance) The Llama Stack project from Meta provides standardized APIs, component abstractions, and integrations with other open source tools and managed services to help develoeprs build and deploy AI applications and agents.

We are organizing several reference implementations of agents on Llama Stack in the Llama Stack Lab part of the Open Agent Lab.

Evaluation and Safety

How do we know that applications built with AI are trustworthy, that they perform as required, in particular that they are safe, free of harmful outputs?

Trust actually has a broad interpretation. Increasingly, organizations moving from proofs of concept to production are not only concerned about evaluating their chosen models and applications for safety, but also for general alignment; do they actually perform well for the specific use cases implemented? This focus area is now exploring these areas of evaluation, in general terms and for specific domains.

Links Description
Trust and Safety Evaluations Initiative AI Alliance icon
TSEI seeks to define the global taxonomy of evaluations (from safety to performance to efficacy), catalog available implementations of them, and provide a reference stack of industry-standard tools to run the evaluations. Our projects:
Ranking AI Safety Priorities by Domain AI Alliance icon
What are the most important safety concerns for your specific domain and use cases? This project explores these questions in several industries, healthcare, finance, education, and legal, with more to come.
The AI Trust and Safety User Guide AI Alliance icon
Introduction to T&S with guidance from diverse experts.
unitxt
Unitxt is a Python library for enterprise-grade evaluation of AI performance, offering the world's largest catalog of tools and data for end-to-end AI benchmarking. (Principal developer: IBM Research)

Deployment and Scaling

Deploying and scaling AI systems, especially to the growing diversity of hardware accelerators for AI, and efficently scaling from PoCs and single node deployments to large numbers of users and distributed deployments are a key set of challenges.

Links Description
The AI Accelerator Software Ecosystem Guide AI Alliance icon
A guide to the most common AI accelerators and the software stacks they use to integrate with tools you know, like PyTorch.

Open Trusted Data and Models

Datasets with clear license for use, backed by unambiguous provenance and governance controls, are needed to train and tune models. A variety of models are needed, not just for English text, but multilingual, multimodal, and domain specific, like models for molecular discovery, geospatial, and time series.

Links Description
The Open, Trusted Data Initiative AI Alliance icon
Open data has clear license for use, across a wide range of topic areas, with clear provenance and governance. OTDI seeks to clarify the criteria for openness and catalog the world’s datasets that meet the criteria. Our projects:
Open Models
Several AI Alliance work groups are collaborating on the development of domain-specific models:
  • Semikong - The world's first open LLM tuned specifically for the semiconductor industry. (Principal developers: Aitomatic, Tokyo Electron Ltd., FPT Software, and The AI Alliance)
  • Llamarine (coming soon!) - An LLM tuned specifically for the needs of the maritime industry.
  • Materials and Chemistry work group (Several developers, including IBM Research):
    • smi-ted - SMILES-based Transformer Encoder-Decoder (SMILES-TED) is an encoder-decoder model pre-trained on a curated dataset of 91 million SMILES samples sourced from PubChem, equivalent to 4 billion molecular tokens. SMI-TED supports various complex tasks, including quantum property prediction, with two main variants (289M and 8×289M).
    • selfies-ted - SMI-SSED (SMILES-SSED) is a Mamba-based encoder-decoder model pre-trained on a curated dataset of 91 million SMILES samples, encompassing 4 billion molecular tokens sourced from PubChem. The model is tailored for complex tasks such as quantum property prediction and offers efficient, high-speed inference capabilities.
    • mhg-ged - SELFIES-based Transformer Encoder-Decoder (SELFIES-TED) is an encoder-decoder model based on BART that not only learns molecular representations but also auto-regressively generates molecules. Pre-trained on a dataset of ~1B molecules from PubChem and Zinc-22.
    • smi-ssed - Molecular Hypergraph Grammar with Graph-based Encoder Decoder (MHG-GED) is an autoencoder that combines a GNN-based encoder with a sequential MHG-based decoder. The GNN encodes molecular input to achieve strong predictive performance on molecular graphs, while the MHG decodes structurally valid molecules. Pre-trained on a dataset of ~1.34M molecules curated from PubChem.
  • More to be announced soon.
Parsing PDFs to Build AI Datasets for Science AI Alliance icon
There is a need for domain-specific datasets for tuning special-purpose models and use in data-heavy application patterns like RAG. In many technical domains, much of the expertise is published in the open, but difficult to exploit as AI training data. For example, while text extraction from PDFs is already common, extracting information from the tables and images in technical PDFs, and associating that information with the surrounding text, is not yet widespread. This project aims to solve this problem by applying the recently-developed Docling tool to parse PDF datasets and create new datasets in formats that preserve this rich content, yet are easier to exploit in model training, tuning, etc. The project will start with the Math-PDF dataset of PDFs published recently by PleIAs.
TerraTorch
TerraTorch is a library based on PyTorch Lightning and the TorchGeo domain library for geospatial data. (Principal developer: IBM Research)
GEO-bench
GEO-Bench is a General Earth Observation benchmark for evaluating the performance of large pre-trained models on geospatial data. (Principal developer: ServiceNow)

For More Information

Alliance image

  • Contributing: We welcome your contributions! Here’s how you can contribute.
  • About Us: More about the AI Alliance and this project.

For More Information

Authors AI Alliance Team
Last Update V0.2.0, 2025-06-14