Port a CPU Monte-Carlo Simulator to GPU for an Energy Trader
Overview
What this challenge is about.
Receive the existing simulator (C++17 + OpenMP, around 5,000 lines), the test book (around 8,000 spread options), and access to an H100 80GB. Port to CUDA: random-number generation with cuRAND, path generation in shared memory with warp-level reductions for pathwise stats, partial sums accumulated with atomicAdd-free tree reduction. Tune block size, registers per thread, occupancy. Validate numerical equivalence: max relative error < 1e-4 on PnL across the 8,000-option book; also verify Greeks (delta, vega) match within tolerance. Profile with Nsight Compute and Nsight Systems. Deliver the CUDA simulator, the validation report (PnL + Greeks), the Nsight reports (annotated), a deployment Dockerfile, and a 6-page handover memo for the quant-tech lead.
The Brief
What you'll do, and what you'll demonstrate.
Port a Monte-Carlo simulator to CUDA, hit >=70 percent peak FP32 throughput on an H100, and validate PnL + Greeks within 1e-4 vs. the CPU baseline.
Earning criteria — what you'll demonstrate
- Port Monte-Carlo kernels to CUDA with warp-level reductions
- Tune occupancy and register usage on an H100
- Validate financial-math equivalence with Greeks, not just PnL
- Package CUDA workloads for deployment to a shared GPU cluster
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Career mappings coming soon.