Research behind the engine

Published papers

Qelly's decision engine uses ideas developed in the papers below — quantum-annealer formulations for probability estimation and portfolio sizing under sparse, probabilistic constraints. Production execution is hybrid quantum-classical; D-Wave QPU routing is in calibration.

arXiv:2605.17628 · 2026

A Penalty-Free Pipeline for Direct Quantum-Annealer Portfolio Optimization

Luis Lozano

arXiv ↗

Penalty-free QUBO encoding reduces chain breaks and improves feasibility on D-Wave portfolio problems.

Reformulates portfolio optimization for D-Wave hardware by eliminating the standard penalty encoding. The penalty-free pipeline yields lower chain-break fractions and more feasible solutions on benchmark instances, with results that connect directly to how Qelly's hybrid optimizer is structured on sparse, constrained event-market portfolios.

arXiv:2605.17623 · 2026

Where the Quantum Lives in D-Wave Hybrid Portfolio Optimization

Luis Lozano

arXiv ↗

Audit of D-Wave's hybrid portfolio service: QPU contribution is minimal; classical post-processing dominates.

Audits the D-Wave Leap hybrid portfolio service end-to-end and measures the actual QPU contribution to final solution quality. The QPU's share of useful work is small (~0.7%) and most of the gain comes from classical pre/post-processing. This is the evidence behind Qelly's honest framing: hybrid quantum-classical optimization, with the D-Wave QPU calibrating into production routing.

How this shows up in Qelly

Every Live Lab decision is tagged with its optimizer backend (Classical / Quantum (CPU) / D-Wave QPU / Manual). The methodology page explains how each backend is used and what's reported on /live-lab.

Reported metrics are research diagnostics on a small, company-funded portfolio. Not a performance claim.