AI-augmented fund valuation pipeline with dual anomaly detection (Isolation Forest + Autoencoder), yield curve pricing, Level 3 DCF valuation, and multi-currency support.
Live Treasury yields feed spread-based pricing for bond ETFs. Each bond has an assumed credit spread and duration; yield shocks translate to price impacts via ΔP ≈ -D × Δy.
Architecture: 3 → 8 → 1 → 8 → 3 (PyTorch). Trained 200 epochs, Adam lr=0.005. Reconstruction error above the (1-contamination) percentile = anomaly. Consensus flag = EITHER model flags.
FV = Σ[FCF_t/(1+r)^t] + TV/(1+r)^n. Terminal value: exit multiple (PE/RE/distressed) or par recovery (credit). 5×5 sensitivity heatmaps stress-test ±200bps discount rate × ±300bps growth rate.
GBP/EUR positions converted to USD via live FX rates. All NAV calculations in USD-equivalent.