Decoding Decay - Real-Time Particle Detection from Point Clouds

Context

Dive into the heart of particle physics with Belle II, located in Tsukuba, Japan. In our quest for indirect evidence of dark matter and other rare decays, we are striving to improve the detector’s efficiency through point cloud neural networks. Recently, these algorithms have shown promising performance in classic computer vision applications.

In this thesis, you will discover what it takes to adapt point cloud algorithms for online processing in large-scale scientific experiments such as Belle II. You will have the opportunity to gain experience in FPGA firmware development, embedded system integration, and algorithm-hardware co-design for high-performance computing systems.

Objectives

You will work with us to improve the performance of our state-of-the-art point cloud network accelerator using top-of-the-line FPGAs and CGRAs through innovative digital design automation methods. Together, we will bring real-time machine learning on FPGAs for particle physics one step closer to reality.

Requirements 

• We’re looking for enthusiastic students who are eager to learn and grow with us.
• Ideally, you have taken one of these courses: System-on-Chip Laboratory (PSoC), Hardware Design Laboratory (DHL), Hardware Synthesis and Optimization (HSO), Hardware-Software Co-Design (HSC), or Digital Circuit Design (DDS).
• Experience with FPGA firmware development is a plus, but not required.
• You should enjoy working in a diverse, international, and interdisciplinary team.
• You must be fluent in either German or English.