WISER Global Quantum+AI 2026 · Vanguard Challenge

Feasibility by Physics,
Not by Penalty

A penalty-free hybrid quantum–classical pipeline for multi-asset portfolio construction.

The idea

Pick exactly K assets from a large universe and allocate capital across them — without ever breaking a hard institutional guardrail. The standard quantum approach encodes "exactly K" as a QUBO penalty you tune with a knob; a penalty only encourages the rule and can silently break it.

Measured results

Every number is reproducible from the committed code — N = 100, K = 20, seeded.

0
hard-guardrail breaches, guaranteed by design
1.0000
approximation ratio vs certified optimum
1,735
routed 2-qubit gates vs the penalty method's 24,530
0
breaches even under ±20% return-forecast error

The allocation copilot

An interactive Streamlit dashboard — shown here on 41 real multi-asset ETFs. (This page is a static showcase; the live app runs from the GitHub repo.)

Allocation tab: weights, metrics, investor-goal controls
Allocation. Final book on real ETFs — zero breaches, holdings pinned at the position cap and floor, four investor-goal sliders.
ADMM convergence residual charts
ADMM convergence. Recorded primal/dual residual trajectories and the adaptive-ρ trace — the classical/quantum handshake, plotted as evidence.
Heavy-hex transpilation audit table and chart
Hardware audit. Real transpiler runs onto a heavy-hex lattice — OMEGA's largest circuit vs the penalty method's monolithic one.

This is a simulator-based research demo (statevector / tensor-network), not real-QPU execution. At these cluster sizes classical enumeration is exact and the quantum layer matches it — the contribution is the architecture and the compliance guarantee.