Skip to main content
School of Physical and Chemical Sciences

Data-Centric Engineering in Particle Physics: Using Machine Learning for Particle Selections on Fast Custom Electronics

Research Group: PPRC
Number of Students: 1
Length of Study in Years: 4 years
Full-time Project: yes


This project is only available to applicants who qualify for the Mexican Conacyt program

  • Applicant required to start in September 2024.
  • The studentship arrangement will cover overseas tuition fees for the duration of the studentship.

Project Description

The ATLAS experiment at the Large Hadron Collider collects proton-proton collision event data at the highest energies ever achieved on Earth. In this PhD project state-of-the-art machine learning algorithms will be developed to rapidly filter the live data searching for new physics.

At the LHC collisions occur every 25 nanoseconds (or at a 40 MHz collision frequency) generating extremely large data sets. Particle physicists are interested in the rare collisions at the very highest energies which include the production of the Higgs boson and signatures of new physics such as the presence of new particles. The majority of the collision data are low energy interactions which are not of interest and must be carefully filtered. The ATLAS detector has a 2-level trigger and data-acquisition system to filter this data. The first level is implemented on fast custom built pipelined electronics filtering data from 40 MHz down to 100 kHz. The second layer is implemented in software to further reduce the rate to 1 kHz, selecting only the rare collisions of interest.

You will develop and test new machine learning approaches to select events in which there are high energy tau particles produced, or the production of high momentum electrons. The trigger electronics are pipelined in buffers such that the online selection algorithms have about 2 microseconds to make a decision based on the readout data from the Level-1 Calorimeter Trigger. The second phase of the project will then perform an analysis of the data using advanced statistical techniques to search for signatures of new physics.

A comprehensive training programme is offered involving 150 hours of lecture courses in particle physics and detection methods, a 2-weeek residential summer school, and attendance at an international conference to present your work towards the end of the PhD. Students may also participate in the DISCnet training programme in machine learning, or have the opportunity to spend 6-12 months based at CERN, in Geneva depending on additional funding.

The project will be supervised by Prof Rizvi who has an established expertise in machine learning and operating the Level-1 Calorimeter Trigger. He leads the ATLAS group at Queen Mary University of London, and has successfully supervised 15 PhD students, and currently leads two Centres for Doctoral Training in Data-Centric Engineering and in Data-Intensive Science for Particle Physics. He is a Fellow of the Alan Turing Institute – the national centre for Artificial Intelligence.


Applicants must be eligible for the Mexican Conacyt programme.

  • Applicant required to start in September 2024.
  • The studentship arrangement will cover overseas tuition fees for the duration of the studentship.

SPCS Academics: Professor Eram Rizvi