Demo Day Presentation & Portfolio

Updated

October 28, 2026

Demo Day Presentation & Portfolio

Presenting your work effectively is as critical as writing fast code.

Deliverables Checklist

  1. GitHub Repository: Clean, modular repository containing your kernel source, CMake / setup.py build configuration, and automated test runners.
  2. Performance Roofline Graph: A generated plot showing where your kernel sits relative to theoretical memory and compute ceilings across varying matrix dimensions.
  3. Speedup Comparison Table: Side-by-side throughput (TFLOP/s) comparisons against naive CUDA, PyTorch torch.compile, and official vendor libraries (cuBLAS / cuDNN).
  4. Technical Defense: Prepare concise explanations for architectural decisions: tile sizes, shared memory allocations, register counts per thread, and how AI-assisted iteration shaped the final solution.