hw-201
Intermediate
Available
An engineer's tour of the modern data center GPU: how SMs and CUs, warps and wavefronts, tensor and matrix cores and the on-chip memory hierarchy work; what FP64 through FP4 and sparsity really buy you; how HBM, PCIe, NVLink, Infinity Fabric, UALink and CXL move data; and how power, cooling and reliability features shape real deployments. Reviewed against vendor sources as of October 2026.
hw-202
Intermediate
Waitlist
How server CPUs are built and where they matter in AI systems: cores, chiplets, caches, memory channels and PCIe lanes; vector and matrix extensions from AVX-512 and AMX to SVE2 and SME; the 2026 Xeon 6, EPYC 9005 and 9006, Grace and Vera, and cloud Arm lineups; and how to pick and tune a host CPU so GPUs are never starved. Reviewed against vendor sources as of October 2026.
hw-203
Intermediate
Available
The physical layer of AI networking: SerDes and lane rates, NICs and SuperNICs, DPUs, InfiniBand and Ethernet switch silicon, rail-optimized topologies, and the transceivers, cables and co-packaged optics that connect them. Covers ConnectX-8 and ConnectX-9, BlueField-3 and BlueField-4, Quantum-X800, Spectrum-X, Tomahawk 6, Pollara and Thor Ultra, and Ultra Ethernet, reviewed against vendor sources as of October 2026.
hw-204
Intermediate
Available
The hardware view of AI storage: NAND types and SSD internals, endurance and DWPD, NVMe and EDSFF form factors, PCIe Gen5 and Gen6 drives, high-capacity QLC, storage servers and arrays, NVMe over Fabrics, GPUDirect Storage and KV-cache storage tiers, and how to size bandwidth and capacity for training, checkpoints and inference. Reviewed against vendor sources as of October 2026.
hw-205
Intermediate
Waitlist
Every memory tier an AI system uses: DRAM fundamentals, HBM3E and HBM4 stacks and packaging, DDR5 RDIMM and MRDIMM, LPDDR5X and SOCAMM modules, CXL memory expansion and pooling, and memory reliability. Ends with capacity and bandwidth planning for weights, KV cache, activations and optimizer state. Reviewed against JEDEC, CXL Consortium and vendor sources as of October 2026.
hw-206
Intermediate
Waitlist
The silicon behind on-device AI: NPU, GPU and DSP architectures, what TOPS ratings do and do not mean, the NVIDIA Jetson Orin and Thor families, PC and mobile SoC NPUs from Qualcomm, Apple, Intel, AMD and MediaTek, discrete edge accelerators, and the quantization, toolchain, power and thermal work needed to ship a model on a device. Reviewed against vendor sources as of October 2026.
hw-207
Intermediate
Waitlist
The physical stack under embodied AI: robot architectures and form factors, onboard compute such as Jetson AGX Thor, cameras, depth sensors, LiDAR, IMUs and tactile sensing, motors, gearboxes and grippers, real-time buses like EtherCAT and CAN, batteries and power, and the functional-safety standards that govern robots working near people.
hw-301
Expert
Waitlist
An architect's comparison of AI accelerators that are not general-purpose GPUs: systolic-array TPUs, AWS Trainium and Inferentia, Intel Gaudi and its successors, Cerebras wafer-scale engines, SRAM-centric dataflow and LPU designs from Groq and SambaNova, and hyperscaler in-house chips such as Maia and MTIA. Covers microarchitecture, memory and interconnect, compilers and software stacks, and a structured method for evaluating them against GPUs. Reviewed against vendor sources as of October 2026.
hw-302
Expert
Waitlist
How AI servers and racks are built: the HGX baseboard and 8-GPU node, MGX modular servers, rack-scale systems such as GB300 NVL72, Vera Rubin NVL72 and AMD Helios, OCP Open Rack v3 and Open Rack Wide, power delivery from 48 V busbars to 800 VDC, direct liquid cooling and CDUs, management and serviceability, and how to plan, procure and commission them. Reviewed against vendor and OCP sources as of October 2026.
hw-303
Expert
Waitlist
A practitioner's method for proving AI hardware healthy and finding out why it is not: acceptance criteria, layered burn-in from component to fabric, GPU diagnostics with DCGM and the ROCm Validation Suite, CPU, memory, storage and network stress, silent data corruption screening, telemetry and error codes, failure analysis and RMA practice. Reviewed against vendor documentation as of October 2026.