Objectives & Outcomes Primary goals: • Demonstrate market microstructure understanding via synthetic LOB generation and execution stress. • Produce cross‑sectional alpha using a graph‑structured learning approach (GraphSAGE/GATv2). • Formulate mean–variance selection as QUBO and solve with QAOA; compare to classical optimizers. • Ship an end‑to‑end backtest and a dashboard; Key outcomes: metrics tables, ablations (with/without GAN augmentation; GNN vs linear), and deployable code.