What survived the noise, written down. The full record keeps itself on ORCID and INSPIRE.
Rejecting beam-induced background on the detector itself at a 10 TeV muon collider — muon decays in the ring flood the detector, so the filtering has to happen before the data ever leaves the front end.
Probabilistic inference implemented directly on-chip at the sensor edge for charged-particle tracking — pushing AI/ML into the detector front-end under extreme latency and power constraints.
An end-to-end tool for hardware-aware compression of machine-learning models — structured pruning and quantization tuned for deployment under real-time, resource-constrained conditions.
Modeling quasi-periodic oscillations in accreting compact sources as a nonlinear hydrodynamical resonance, and using them to constrain the spin of the central compact object.
Robustness of the smartpixels classifier for different simulated sensor geometries and non-ideal detector conditions
Nucl. Instrum. Meth. A 10922026
Performance measurements of the electromagnetic calorimeter and readout electronics system for the DarkQuest experiment
Nucl. Instrum. Meth. A 10802025
Sensor Co-design for smartpixels
arXiv preprint2025
Characterization of a 28 nm smartpixels ASIC With On-Chip ML for Particle Tracking Detectors
arXiv preprint2025
Asymptotically flat vacuum solution for a rotating black hole in a modified gravity theory
Eur. Phys. J. C 822022
Scaling Out hls4ml for ASICs: A Surrogate Model for Neural Network Accelerator Synthesis Metrics
September 2026