Overview
What this challenge is about.
Receive the existing pricer (C++17, scalar, around 600 lines), 2 weeks of representative input data (around 50M paths), the validation harness, and access to an Intel Ice Lake-SP test box. Re-architect the pricer to process 16 doubles per AVX-512 vector lane, mind the special-function (erf, exp, log) calls — either use Intel SVML or hand-roll polynomial approximations. Build a portability variant using Highway so the code also runs on ARM SVE. Benchmark with perf + Google Benchmark across path counts from 1M to 50M. Validate numerical equivalence: max relative error < 1e-6 across the full dataset. Deliver the rewritten pricer (both intrinsics + Highway versions), the benchmark dataset, the numerical-validation report, a 6-page performance memo for the risk-tech head, and a Compiler Explorer link demonstrating the generated assembly.
The Brief
What you'll do, and what you'll demonstrate.
Vectorize a Black-Scholes-heavy VaR pricer to fit a 90-minute window using AVX-512 (with a portable Highway variant) while staying numerically equivalent within 1e-6.
Earning criteria — what you'll demonstrate
- Apply AVX-512 intrinsics to a real numerical kernel
- Write portable SIMD via Highway without sacrificing performance
- Validate numerical equivalence under vectorization
- Benchmark with statistical rigor (Google Benchmark, perf)
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.