Lekton.AI

Research & Innovation

Where we look, beyond the day-to-day

Our products live in the present of generative AI: document platforms, compliance, operational agents. But part of our work looks further ahead. In collaboration with the University of Siena we pursue a line of research into advanced frontiers of artificial intelligence, with a common principle: rigorous validation, declared uncertainty and open code wherever possible.

Quantum Machine Learning

We explore hybrid quantum-classical neural architectures and new algorithms for the efficiency of quantum circuits, on the Qiskit and PennyLane platforms. The opening of this research front is described in the article Lekton.AI opens a new research front: Quantum Machine Learning.

  • qh_algorithms — a rigorous experimental comparison between a classical CNN, a hybrid quantum-classical CNN (HQCNN) and a purely quantum reference, with multi-seed statistics, Wilcoxon tests and a noise-resilience study. MIT licence.
  • QC-2025-26 — the notebooks of our Quantum Computing lab: from the circuit library to the Deutsch-Jozsa, Grover, Quantum Phase Estimation and Shor algorithms, on simulator and real hardware.

AI for healthcare and medical imaging

Classical deep learning applied to clinical images, with attention to methodology and to the protection of health data. The case we have described combines classical and quantum AI on the early diagnosis of Parkinson’s disease: Not just generative AI: classical and quantum AI in the service of Parkinson’s disease diagnosis.

  • BrainNet — a 3D convolutional neural network for the classification of C-DOPA PET images, with a leak-free pipeline, patient-grouped cross-validation and metrics reported with uncertainty.

The AI-for-healthcare projects are research activities: they do not constitute medical devices or diagnostic tools validated for clinical use.

Academic roots

Much of this research is linked to the doctoral thesis “New Perspectives on Quantum Technologies: Progress on Quantum Sensing and Quantum Computation” (University of Siena, Cycle XXXVII). The repositories qh_introduction, qh_algorithms, qh_hardware and qh_sensing collect the code and figures of the chapters, with the LaTeX sources in PhDThesis.

All the research code is public on our GitHub. Want to collaborate or find out more? Contact us.