The course for engineers building the infrastructure that ML runs on. Covers distributed training communication (NCCL, MPI), custom CUDA kernels for attention and matmul, model parallelism strategies, and building a parameter server in C++. Used by engineers at ML infrastructure teams.
Duration
20 hours
Students
480+
Rating
⭐ 4.9 (9)
After enrolment you will receive access instructions by email within 24 hours. Course materials are delivered via our online learning portal.
“Practical, rigorous, and immediately applicable. Velmio courses are genuinely different.”
Mark T.
Head of Data
Order Summary
C++ for ML Infrastructure
By enrolling you agree to our Terms & Conditions and Privacy Policy.