New! NDIF is opening up more spots for the 405b Pilot Program. Apply now!
New! Applications for NDIF's Engineering Fellowship are now open. Apply now!
Get Involved
NDIF is under continuous development, and there are several ways to get involved in the prototype fabric during its construction.
- Join our community of early adopters. Our online discord server hosts students and researchers from many fields who are dissecting the behavior and mechanisms of large AI models. Request early access to the prototype NDIF service through this community.
- Use our open-source library, nnsight. NNsight can be used with NDIF and it can also be used with your own local computation power, independently of the fabric. It works with pytorch and can be easily installed via pip.
- Join our team. We are looking for talented and motivated team members who are inspired to create a vibrant scientific community to crack the mysteries of large-scale AI. Jobs and opportunities are listed on this page.
The NDIF project code is open-source and can be found at github.com/ndif-team.
An NDIF workshop for undergraduate and graduate students, introducing state-of-the-art methods for performing interventions on the internal computations of large language models, using NDIF. NDIF enables students and scientists to share GPU resources to learn, develop, and deploy scientific methods that crack open the internals of large neural networks
NDIF Community
On the NDIF community discord, you can chat with the team and discuss tips, tricks, and the latest research. Submit a short form to join.
Join the groupNNsight
You can use NDIF right away in local mode without an account by using the nnsight library. Just a "pip install" away. Read about it here.
Go to nnsightJobs
We are looking for help building NDIF. To join the NDIF team full-time, part-time, as a co-op or a volunteer, see our job listings on this page.
See jobs belowJobs and Community Opportunities
We are seeking a highly skilled Principal Research Software Engineer with experience in Machine Learning and Large Language Model interpretability research methods, to assist in developing the National Deep Inference Fabric, an open-source deep learning interpretability research computing infrastructure project.
You will be responsible for full stack development, doing both back-end and front-end software development to help create a robust, high-throughput, highly usable, and flexible multi-tenant AI inference service to enable research nationwide. Some of the day-to-day activities include solving security, stability, integration, and performance issues involved in providing a large-scale research inference service for open-source AI models.
We are looking for someone who can implement state-of-the-art parallel GPU inference methods, and incorporate them into a system with job scheduling, routing, quota management, authentication, authorization, and telemetry to create a high-performance computing infrastructure. This person should be expert in Python and working internals of PyTorch along with Unix/Linux service development, HPC/cloud environments, and all other aspects of the software development life cycle.
This is a full-time role, based in Boston, MA
The Summer 2025 NDIF Engineering Fellowship is an intensive program for engineering students to contribute to an active state-of-the-art AI research engineering project in the public interest. The program welcomes applications from PhD, Masters, and undergraduate students in computer science who have knowledge and interest in ML Systems or full-stack application development.
In the future we will post specific listings for other roles we seek to fill including
- Student research assistants.
- Student co-ops.
We also welcome research collaborators and unpaid open-source contributors; for open-source community opportunities, get in touch through our community Discord.