IBM Quantum Developer Conference 2026 · Chicago

HADOF

Industrial Scale Applications on Today’s QPUs

Namasi G. Sankar¹ · Georgios Miliotis² · Simon Caton¹

¹ School of Computer Science, University College Dublin
² University of Galway

The Hamiltonian Auto-Decomposition Optimisation Framework splits problems far larger than any QPU into circuits that fit on it.

The gap

Real problems are hundreds of times bigger than the chip

45,000

variables in our largest hardware run

1.01 bn

pairwise couplings in that one problem

156

qubits available on the IBM device

How it works

Solve a few qubits at a time

Decompose Split into groups that fit one circuit

Solve on the QPU The rest enters as probabilities

Share and sweep Samples update every probability

Recombine Keep the best full solutions

Toy run: 24 variables in 3-qubit groups. Shade = probability a variable is 1; the red outline is the group on the QPU.

Proven on IBM hardware

Small circuits hold their quality

Best quality vs simulated annealing · ibm_kingston

3–4×

less wall-clock time across four IBM QPUs

0.98

of SA’s quality at 500 variables with 5-qubit circuits

Flagship application

Quantum genome assembly of the P. aeruginosa bacterial chromosome

Exploiting today’s quantum computers for real optimisation workloads.

Reads

21,969

real nanopore reads

QUBO

2,313

variables, >17× the qubits

ibm_torino

771

3-qubit circuits per sweep

Chromosome

7.14 Mbp

circular, polished

99.348%

of the genome recovered on IBM ibm_torino, duplication ratio 1.0

To our knowledge, the first complete genome assembled from real sequencing data on quantum hardware: over 100× larger than earlier hardware demonstrations.

New · Topology-aware HADOF

From 5-qubit to 150-qubit circuits

Published HADOF

5 qubits per circuit

Topology-aware · ibm_aachen

150 qubits per circuit

Random dense QUBO

22,350 vars

45,000 vars

QAOA-HADOF on ibm_aachen

52.6%

49.1%

SA-HADOF, same groups

98.5%

98.8%

SA-HADOF speed-up vs full SA

7.9×

8.6×

QPU execution time

~35 min

73 min

Time waiting in the queue

~6.4 h

~74.6 h

Quality = best objective as a share of full simulated annealing on the same QUBO.

Method unpublished. Ask me how.

Next: a genome ~100× larger

Runs any QUBO

with any sampler: QAOA, quantum annealing, SA, FALQON

Genome assembly · on hardware

Dense QUBOs · to 45,000 vars

Routing and logistics

Scheduling

Portfolio optimisation

Network design

Feature selection

Outbreak and cancer genomics

Sankar, Miliotis & Caton · AIQxQIA @ ECAI 2025, CEUR-WS 4153 · arXiv:2604.27836 · Quantum genome assembly on NISQ hardware (2026). ibm_aachen time funded by the IBM Quantum Credits programme; PhD funded by a UCD School of Computer Science scholarship.

See it run

Live, animated version with links and papers

namasi-phd.github.io/qdc2026
github.com/Namasi-PhD
namasivayam.gomathisankar@ucdconnect.ie