Capstone Projects
Capstone project topics for 2026/27
Quantum AI, responsible AI, and benchmarking topics designed for undergraduate capstones with clear deliverables.
Topics for capstones 2026/27
Below is the shortlist of proposed capstone topics. Each topic is designed to support (i) a clear research question, (ii) a reproducible experimental plan, and (iii) a tangible deliverable (code + report). Students are free to propose other topics within my areas of expertise.
Quantum for image classification (MedMNIST)
Evaluate quantum and hybrid quantum–classical classifiers on MedMNIST (18 datasets), with fair baselines and robustness analysis.
- Baselines: CNN/ViT + classical ML
- QML: VQC/QNN, quantum kernels
- Ablations: embeddings, depth, noise, compute cost
Quantum for image segmentation (MRI / X-ray)
Investigate hybrid segmentation pipelines (e.g., U-Net + quantum layers/feature maps) and assess performance vs complexity.
- Dice/IoU + failure mode analysis
- Hybrid approaches and scaling limits
- Reproducible training/evaluation code
Quantum for tabular data (time series)
Apply QML to structured data and time series, studying encodings, sample efficiency, and generalization under constraints.
- Angle/amplitude/learned embeddings
- Limited data and noisy labels
- Compute and stability analysis
Quantum for multimodal data
Explore quantum or hybrid fusion for multimodal clinical data (images + reports + labs), including missing-modality robustness.
- Fusion strategies with/without quantum components
- Calibration and uncertainty
- Robustness to missing modalities
Quantum federated learning for healthcare
Prototype privacy-preserving federated learning with quantum/hybrid models—focus on threat models, privacy leakage, and mitigations.
- Federated training prototype (simulation or multi-client)
- Privacy/security discussion and safeguards
- Performance vs privacy + communication cost
Responsible AI & Ethics: clinical documentation revealed by LLMs
Study memorization, leakage, and re-identification risks in clinical NLP workflows, with practical guidelines for safe deployment.
- Risk taxonomy and literature review
- Reproducible evaluation (approved or synthetic data)
- Mitigation: redaction, auditing, access control
Quantum finance (SAT / hybrid optimization)
Explore hybrid pipelines where part of a finance problem is mapped to SAT/QUBO and executed on quantum hardware/simulators.
- Formulation + classical baseline solvers
- Hybrid quantum approach (e.g., QAOA)
- Scaling and quality trade-offs
Benchmarking quantum AI algorithms 2 students
Build a rigorous benchmarking framework for QAI across tasks and datasets with standardized metrics and compute budgets.
- Experiment harness + baselines + reporting
- Noise studies + simulator/hardware integration
- Documentation + CI + generated results page
How to apply: email a short statement of interest, transcript, and CV to one of the contacts below.