About the team
Scale the models everyone is talking about. Today's models span hundreds of GPUs, and the network between them decides whether a training run takes a week or a month.
As AI models became larger and more complicated, a performant graph mode and efficient scaling solution became a key aspect of the SW solution.
Our team builds that layer. We own the Intel oneAPI Collective Communications Library (oneCCL) and develop the software for multiple generations of Intel AI GPUs, for training and inference.
What you'll do
•You'll be a member of the team, working on the library that moves data between GPUs in some of the largest AI training jobs running on Intel hardware.
•Develop features in the core library: design and implement parts of the C++/SYCL code that run on the GPU, from the first design through code review to the release.
•Push performance: profile communication across many GPUs, find the bottlenecks and make the fixes that cut the time it takes to train AI models.
•Solve hard bugs: track down concurrency, synchronization and memory issues in a distributed system that runs across many GPUs at once.
•Shape how the team builds software: improve the build system, CI pipelines and developer tooling the whole team depends on every day.
•Learn to use AI tools as part of the development process.
What you'll get
•Hands-on work with code that runs in production and is open source.
•A mentor from the team and regular one-on-one guidance.
•Experience with high-performance computing and AI infrastructure, a fast-growing field.
Qualifications
Required
•B.Sc. or M.Sc. student in Computer Science, Computer/Electrical Engineering or a related field.
•Good working knowledge of C++ from courses or personal projects.
•Solid basics in data structures and algorithms.
•Comfortable with, or eager to learn, working in Linux and on the command line.
•Curious, willing to ask questions, and able to learn new things on your own.
•Available 3 days per week.
Nice to have
•A course or project in parallel programming, operating systems or computer architecture.
•Exposure to GPU programming (CUDA, SYCL, OpenCL) or MPI.
•Python scripting.
•Experience with Git and GitHub.
You don't need prior industry experience. We care more about how you think and learn than about what you already know.
Job Type: Student / Intern
Shift: Shift 1 (Israel)
Primary Location: Israel, Petah-Tikva
Additional Locations: Israel, Haifa
Work Model for this Role: This role will require an on-site presence.
