Peter Ma

Hi there! 👋 I'm an incoming first year Astrophysics PhD student at UC Berkeley. Previously, I did my undergrad at the University of Toronto in Applied Mathematics. Broadly, I'm interested in intelligence, finding it, creating it and using it to solve hard problems (physics). I'm fascinated both by the world of atoms and bits.

\(\Rightarrow\) Currently, I am working on applying deep learning to study Dark Matter effects on stellar streams in simulations at the Dunlap Institute of Astronomy and Astrophysics.

In the past I developed ML models and firmware on FPGA's for the High-Luminosity Large Hadron Collider at CERN. Before that, I developed deep learning algorithms to assist controlling LIGO at Caltech. I've also spent time at UofT's Computer Science Dept looking at using transformers for equation discovery and symbolic regression. Before that I developed an ensemble learning algorithm for Fast Radio Burst detection deployed on the CHIME radio telescope at the Dunlap Institute of Astronomy and Astrophysics.

My first research project was developing an end-to-end deep learning algorithm to search for signs of advanced life beyond Earth with Breakthrough Listen.

When I'm not busy teaching computers, I love climbing and making art. Come say hi \(\Rightarrow \)Twitter @peterma02 Email: peterxy.ma [at] gmail [dot] com :)

First Author Published Papers (Undergrad)

A deep-learning search for radio technosignatures from 820 nearby stars
Ma, P. et al. (2023) Nature Astronomy - Published here

A Deep Neural Network Based Reverse Radio Spectrogram Search Algorithm
Ma, P. et al. (2024) RAS Techniques and Instruments - Published here

A Deep Learning Technique to Control the Non-linear Dynamics of a Gravitational-wave Interferometer
Ma, P. & Vajente, G. (2024) Classical and Quantum Gravity IOP - Published here

Towards Identifying the Fundamental Properties of Dark Matter through Graph Neural Network Analysis of Stellar Streams
Ma, P. et al. (2024) - In prep ...

Co-Authored Papers

AstroCompress: A benchmark dataset for multi-purpose compression of astronomical data
T. Truong*, R. Sudharsan*, Y. Yang, P. Ma, R. Yang, S. Mandt, J.S.Bloom (2024) - Submitted Neurips Dataset

Unpublished Manuscripts

Developing Firmware and Algorithms for the Liquid Argon Signal Processor
Ma, P. et al. (2023) - CERN report here

Machine Learning and Simulation Strategies To Improve Fast Radio Burst Detection
Ma, P. et al. (2022) - report here

UofT Dept. Computer Science

Researcher Assistant: Sept 2023 - April 2024

Investigating the emergence of sparsity in the development of deep learning models in solving physics based problems. Also look at the use of Large Language Models and prompting for novel knowledge generation in physics based problems. Supervised by Prof. Vardan Papyan.

UofT - Dunlap Institute for Astronomy & Astrophysics

Researcher Assistant: Sept 2023 - April 2024

First we look into building faster emulators for dark matter simulations. Look into using JAX for faster gradient computation when producing forward progating models. Secondly we look symbolic regression on deep learning models in learning cosmological struture formation. Supervised by Dr. Keir K. Rogers.

CERN - ATLAS / Large Hadron Collider

Intern Researcher: July 2023 - Sept 2023

Continuing my work at McGill [below] I worked on building deep learning algorithms on low level FPGA's to reconstruct energy readouts from the Liquid Argon Digital Signal Processor for the ATLAS detector on the future High Luminosity Large Hadron Collider. I also help around with developing firmware on digital electronics.

McGill University - Experimental Particle Physics Group

Intern Researcher: May 2023 - July 2023

I work on developing high performance FPGA firmware for the Liquid Argon Signal Processing Unit (LASP) to be attached on the ATLAS detector. I also worked on developing functional hardware test for our LASP digital electronics. Supervised by Prof. Brigitte Vachon

California Institute of Technology - LIGO

Intern Researcher: June 2022 - Sept 2022

I work on tackling non-linear dynamic control problems using deep learning. I specifically investigated attention-based state estimators and reinforcement learning for LIGO's locking acquisition to help detect gravitational waves. I work with Dr. Gabriele Vajente on this ambitious project.

Our preprint paper can be found here

UofT - Dunlap Institute for Astronomy & Astrophysics

Researcher Assistant: Dec 2021 - Apr 2022

Work on developing novel ensemble machine learning models for FRB detection with the CHIME/FRB project. My algorithm now actively runs in production to help improve the core detection pipeline. I am grateful to have worked under Prof. Bryan Gaensler! The goal is to one day use these detections to help astronomers answer important questions regarding the origins of these objects and potential cosmological questions.

UC Berkeley SETI Research Center

Co-mentor: Sept 2022 - Dec 2022

I co-mentor with Dr. Steve Croft. a group of undergraduates from UC Berkeley on developing a deep learning based "reverse image search" method for radio spectrograms leveraging the techniques from computer vision.

Intern Researcher: [June 2020 - April 2022] and [Sept 2022 - May 2023] On pause

Currently developing attention-based geometric deep learning models for the MeerKAT telescope to conduct the largest search effort for signs of life beyond Earth, surveying 1 million stars over a span of 2 years. I am supervised by Dr. Cherry Ng and Dr. Steve Croft,.

Previously I explored how deep neural nets like Disentangled B-VAE's can search 820 stars for technosignatures. My first paper was published in Nature Astronomy! Supervisors were Dr. Cherry Ng, and Dr. Andrew Siemion

In high school I helped build a distributed cloud computing platform for Astronomy Research with Dr. Steve Croft and Yuhong Chen!

AI For Good - Volunteering

Volunteer Researcher: May 2022 - Aug 2022

I work on developing Natural Language Processing models for text classification in building the sustainable development goal (SDG) data catalogue pipeline. The goal is to help build tools that power data driven policy making in achieving the 17 SDG goals set by the UN.