Research
Quantum algorithms, optimization, and learning
I study quantum and hybrid approaches to learning and optimization, with an interest in how problem structure, noise, and computational resources affect their performance.
Research interests
Quantum algorithms
My interests include variational and hybrid algorithms, quantum linear algebra, and quantum simulation. I am interested in how data encoding, circuit design, and measurement influence the usefulness of an algorithm.
Quantum optimization
I am interested in quantum and hybrid methods for discrete and continuous optimization, including QUBO and Ising formulations, QAOA, variational optimization, and quantum annealing. I examine solution quality and resource requirements alongside classical alternatives.
Quantum machine learning
My work includes quantum and hybrid learning models, structured neural networks, and low-rank methods. I am interested in regression, classification, reservoir learning, and how representation and measurement choices affect training and generalization.
Current research directions
Structure-preserving quantum–classical models
Most hybrid quantum–classical models flatten images, sequences, and measurements before learning begins, then spend scarce qubits, shots, or parameters rediscovering that structure. I place small trainable quantum circuits inside compressed tensor bottlenecks, so their cost follows the bottleneck width instead of the raw input, and ask the more useful question of when a quantum component helps.
Measurement-aware quantum learning
A well-placed quantum circuit is only as useful as what is measured from it. ML-QRC preserves the time–qubit–observable structure of reservoir measurements, reducing readout parameters by about 12.1× without sacrificing predictive quality. VQBR separates a regression vector into a direction recovered by a shallow circuit and an analytically solved scale, recovering the MAP direction with 0.998 cosine similarity under finite-shot noise. AdaRO-Q extends this view to task-incremental quantum continual learning.
VQBR · IEEE QCE 2026ML-QRC · QML@QCE 2026AdaRO-Q · in preparation
Tensor methods and quantum optimization
In MTL and its adaptive-rank extension, late convolutional blocks are replaced by mode-wise multilinear transformations and the effective rank is learned rather than fixed by hand. For tensor decomposition, I formulate adaptive CP-rank selection as an interaction-aware QUBO so that a quantum optimizer selects components jointly instead of pruning them by magnitude alone.
MTL · Asilomar 2026LR-MTL · IEEE TAI, under reviewQALR-CPD · QMI, submittedReview · ACM TQC, submitted
Deployable, resource-aware quantum computing
I study where these methods should run once compilation, queueing, shots, mitigation, and classical processing are counted honestly, developing end-to-end cost and quality analyses and resource-prediction models for quantum–classical workloads. Related work recovers deployable classical predictors from quantum-encoded states.
How I evaluate methods
My experiments combine mathematical analysis, classical baselines, reproducible simulations, and quantum-hardware studies. I examine predictive or solution quality alongside the resources needed to obtain it. Depending on the problem, these include qubits, circuit depth, measurement shots, optimization effort, and classical computation.
I work primarily with Python, Qiskit, PennyLane, and PyTorch. My publications and GitHub repositories provide further details on individual projects.