Current projects at SDSC
Anomalous particle showers in muon detector
in collaboration with High Energy Physics Laboratory, EPFL
This project aims to develop a novel analysis pipeline for detecting long-lived particles (LLPs) that decay in the muon chambers of the LHCb detector at CERN. Unlike traditional approaches that focus on reconstructing charged particle tracks close to the collision point, this work explores non-standard detector signatures, such as particle showers occurring in distant regions of the detector. These decays produce highly unconventional patterns, appearing as sparse and irregular clouds of detector hits that are difficult to capture with standard reconstruction methods.
To address this challenge, the project develops advanced models to identify and isolate signal-like patterns of detector hits from background activity. These models are complemented by methods for inferring key properties of the originating particle, including its decay position and energy. Together, these approaches aim to extract meaningful physics information from complex detector data and support new strategies for discovering physics beyond the Standard Model.
Project page coming soon on the SDSC website.
Personalized epidural electrical stimulation
in collaboration with .NeuroRestore (EPFL, CHUV, UNIL) & BMIC group (ETH Zürich) & IT’IS Foundation
Every year, over 250 000 people suffer spinal cord injuries (SCI), often resulting in permanent paralysis of the upper and lower limbs. Epidural electrical stimulation (EES) is a promising neuroprosthetic approach that can help restore motor function. One such system, STIMO, is an implanted device that delivers stimulation to the spinal cord, enabling some individuals with chronic paralysis to regain controlled motion, including walking.
Despite its success, STIMO currently relies on long and complex calibration procedures performed by specialized clinicians. This project supports the development of data-driven tools that can automate and personalize the calibration process. By modeling patient-specific spinal anatomy and motor neuron pathways, the goal is to create interpretable simulations of muscle activation in response to stimulation. These tools are designed to reduce the need for expert intervention and make EES therapies more efficient and accessible to a wider patient population.
Detecting novel drug combinations associated with adverse events
in collaboration with Pharmacoepidemiology group (ETH Zürich)
Adverse drug events (ADEs) caused by harmful drug-drug interactions (DDIs) are a major challenge in healthcare, especially for patients taking multiple medications. While some interactions are known before a drug reaches the market, most potential risks—especially involving three or more drugs—remain hidden due to the sheer number of possible combinations in real-world use.
The MedCare project addresses this challenge by applying machine learning to large-scale healthcare data in order to uncover both known and previously unrecognized interactions. In close collaboration with pharmacoepidemiologists, the project focuses on developing interpretable models that can support the discovery of clinically meaningful drug combinations and help improve patient safety in everyday medical practice.