Open-Source AI / HPC Infrastructure & Education

HiCAIN Build. Operate. Learn.

Two pillars, one mission. Cnuas, an open-source, fully software-emulated OCP Open Rack v3 structured GPU cluster you can run on-premises, paired with open-source academic training through the HiCAIN Academy.

🖥️ Explore Cnuas ↗ 🎓 HiCAIN Academy ↗

student@hicain ~ - perftest on the lab fabric

# Run an RDMA write benchmark on the Cnuas virtual fabric
$ ib_write_bw -d hicain0 --report_gbits gpu-node-02

 #bytes  iterations  BW peak[Gb/sec]  BW avg[Gb/sec]  MsgRate
 65536      1000        198.41           197.06        376184

# Inspect the virtual HCA, looks like real Mellanox to userspace
$ ibv_devices
    device           node GUID
    ------         ----------------
    hicain0        0x1ed5cafe00000001

# That was the HiCAIN Academy "Intro to RDMA" lab.
# All software. No hardware required.

The AI/HPC Skills Stack, Taught Properly

HiCAIN covers the full modern AI/HPC software stack, from GPU programming and high-speed networking to distributed training. Each course pairs theory with hands-on labs you can run on a laptop.

⚡

GPU Computing

CUDA and ROCm programming, memory models, SIMT execution, kernel optimisation, and multi-GPU workloads.

🌐

High-speed Networking

InfiniBand, RDMA/RoCEv2 over Ethernet, DCB (PFC/ETS/ECN), and the OFED stack from verbs up to application transports.

🔄

Distributed Computing

Open MPI collectives, NCCL/RCCL communication patterns, and the practical engineering of multi-node training jobs.

🎓

HiCAIN Academy

Full learning management system, structured courses, modules, lessons, quizzes, progress tracking and certificates of completion.

Visit the Academy ↗

🧪

Cnuas-Powered Labs

Hands-on labs run inside the PacketFive Cnuas Virtual AI HPC Infrastructure, an open-source, fully software-emulated OCP Open Rack v3 structured GPU cluster.

About Cnuas ↗

🌍

Open Source

All HiCAIN training material and the underlying LMS platform are open source under permissive licences, for educators, learners and contributors alike.

Two Pillars of HiCAIN

Infrastructure that lets anyone experiment without a million-dollar lab, and education that builds the future AI/HPC workforce.

🖥️

Core platform

Cnuas, Virtual AI HPC Infrastructure

Open-source, OCP Open Rack v3 structured

Cnuas is a 100% software-emulated AI/HPC datacenter, with virtual GPUs, virtual NVLink-equivalent fabric, virtual RoCE and InfiniBand switches and PCIe-attached virtual NICs, all running unmodified Linux software. It powers every HiCAIN lab, an open-source, fully software-emulated OCP Open Rack v3 structured GPU cluster you can run on-premises.

About Cnuas ↗ GitHub ↗

🎓

HiCAIN Academy, Courses & Training

Open-source learning platform

HiCAIN Academy delivers structured, open-source courses across the full AI/HPC software stack, including CUDA, ROCm, OFED, Open MPI, RDMA, RoCE and InfiniBand. It equips engineers, students and practitioners with the deep skills modern AI clusters demand.

🎓 Explore the Academy ↗ What You'll Learn