code profiling

Struggling with slow AI Python on Linux? This hands-on, Bash-first guide shows how to unlock 2–10× speedups without changing algorithms: install optimized BLAS/LAPACK, verify NumPy/SciPy backends, tune thread counts, and profile hotspots; vectorize or JIT with Numba/NumExpr; control OpenMP/PyTorch parallelism and NUMA/affinity; pick CPU vs GPU wisely. Includes copy-paste commands, micro-benchmarks, and a pitfalls checklist.