Week 35 — Coding Projects

Core

Build the capstone.

Suggested capstone options:

• Tiny graphics engine with compute-based post-processing
• Path tracer with progressive accumulation
• Geometry-processing sandbox with PCA/FFT/filtering
• DSP/image-processing toolkit
• Coding-theory visualizer
• PR-inspired options: 1D/2D histogram (grid) localization engine · Particle filter visualizer with live belief heatmap · Kalman filter / extended Kalman filter toy simulator · Sensor fusion demo combining noisy odometry + distance sensor with GPU-accelerated belief update

Recommended stacks: Metal for Apple-native graphics capstone · Vulkan for explicit engine architecture capstone · CUDA for compute-heavy research-style capstone · NumPy as validation/oracle layer throughout.

• Reading: MBT — whichever advanced chapters align with your capstone · Vulkan Book — engine architecture, advanced rendering, resource systems · CUDA Book — performance, memory hierarchy, advanced kernel tuning.
• NumPy: Keep a CPU oracle/reference path for every major subsystem.
• Stretch: Add test scenes and regression tests. Add offline image comparison or numeric result comparison.
• Verify: End-to-end system works · Major math from earlier weeks appears explicitly in code · Results are reproducible enough for debugging.