Almost every quant model ends in an optimisation: a calibration that fits parameters, a regression that fits a line, a portfolio that picks weights. This skill shows what the solvers actually do, with short checks (numbers you compute, steps you order, concepts you match) and finance examples: why a step size that is too big makes gradient descent diverge, why Adam is the default for noisy objectives, why a convex problem has one answer, what a KKT multiplier prices, how an L1 penalty sets weights to exactly zero and how a linear program splits a large order over time. The capstone is a code lab where you write the projection and the solver yourself and test the result on real stock returns.
