Lai Gui

Embrace complexity.
Curiosity finds the way.
LOADING
. . .
Completed
0

Hello there!

My name is

Lai GUI

I am a graduating Master's student in Physics at EPFL, Switzerland. I contribute to extract rare particle-decay signals from billions-scale noise for CERN's particle-collider experiments, to study the fundamental structure of our universe.

Python C++ AI Agents PyTorch XGBoost Scikit-learn Transformers Monte Carlo Simulation HPC Cluster Computing

What is High Energy Physics

It is the study of matter and forces at their most fundamental level.

At CERN in Geneva, the world's largest High Energy Physics research institute, we smash protons together at nearly the speed of light. Under the immense energy of these impacts, the internal structure of the proton is completely destroyed, creating the particle remnants that fly out from the collision point along distinct trajectories; by studying these paths, we can reverse-engineer subatomic structures and discover the laws that bind them together.

The collider processes 40 million proton collisions every single second, therefore our role demands not only designing a stable experimental facility to withstand a highly irradiated environment, but also deploying advanced data analysis techniques to isolate incredibly rare subatomic signals hidden within a highly chaotic background.

Education

2024 – Present
2024 – Present
EPFL (École Polytechnique Fédérale de Lausanne)
Lausanne, Switzerland
Master of Science in Physics
Averaged result: 5.43/6.00
Institute Website →
EPFL
2021 – 2024
2021 – 2024
Imperial College London
London, United Kingdom
Bachelor of Science in Physics with Theoretical Physics
Averaged result: 77.88%. First Class Honours Degree.
Institute Website →
Imperial College London
2017 – 2019
2017 – 2019
Maranatha Christian School
Melbourne, Australia
Secondary School, Victorian Certificate of Education
ATAR 99.85 (top 0.15% percentile)
Institute Website →
Maranatha Christian School
Nanodegrees
May 2026
Anthropic AI Courses
Anthropic Education
· Claude Code 101 [ Certificate ]
Introduction to Claude Code as an AI-powered CLI, covering core agentic workflows for software development.
· Claude Code in Action [ Certificate ]
Hands-on application of Claude Code across real engineering tasks including multi-file editing, testing, and tool use.
· AI Fluency: Framework & Foundations [ Certificate ]
Conceptual frameworks for understanding large language models, covering training, inference, and prompt design principles.
· AI Fluency: AI Capabilities & Limitations [ Certificate ]
Critical evaluation of AI strengths and failure modes, covering reasoning, hallucination, safety, and appropriate use cases.
Anthropic

Research

Mar – Jul 2026
Mar – Jul 2026
Radiation Damage in SiPMs · Emission Microscopy
Master Project (PDM) · LPHE, EPFL
Signal Processing Peak-Finding Image Analysis
Learn more →
EMMI superimposed emission image
Nov 2025 – Jan 2026
Nov 2025 – Jan 2026
Bs⁰ → τ⁺τ⁻ Selection · Boosted Decision Tree
Specialisation Project · LPHE, EPFL
XGBoost BDT scikit-learn Bayesian Optimisation
Learn more →
6π invariant mass separation between signal and physics background
Sept – Nov 2025
Sept – Nov 2025
SiPM Dark Count Rate · Cryogenic to 100 K
Specialisation Project (TSM) · LPHE, EPFL
Waveform Analysis Poisson Statistics Error Evaluation
Learn more →
FWHM vs overvoltage for all laser intensities
Jul – Aug 2025
Jul – Aug 2025
LHCb SciFi Detector Simulation · Geant4
Summer Internship · LPHE, EPFL
★ Best Poster Prize · EPFL 2025
C++ Landau Statistics Huber Estimation
Learn more →
Cluster intensity MPV vs distance for all SciFi mat designs
Feb – Jun 2025
Feb – Jun 2025
SiPMs Single Photon Time Resolution
Semester Project (TP4b) · LPHE, EPFL
EMG Fitting Uncertainty Analysis Waveform Analysis
Learn more →
FWHM vs over-voltage for all SiPMs
Jun – Aug 2024
Jun – Aug 2024
Radiation Damage Evaluation · Geant4
Summer Internship · J-PARC, Japan
ROOT/C++ Geant4/ICEDUST HPC
Learn more →
COMET CyDet Electronics Room shield radiation dose map
Jan – May 2024
Jan – May 2024
CTH Detector Response · Geant4
Bachelor Project · Imperial College London
Bethe-Bloch Formula Landau/Moyal Statistics Geant4
Learn more →
Bethe-Bloch formula with fitted MPV data points
Aug – Sept 2023
Aug – Sept 2023
Neuromorphic Computing · Time-Series Forecasting
Summer Internship · Imperial College London
★ Results Published · Nature Communications 2025
PyTorch GRU Digital Twins Reservoir Computing
Learn more →
Mackey-Glass forecast from nanomagnetic physical neural network
Jul – Aug 2023
Jul – Aug 2023
Jet Flavour Tagging · Particle Transformer
Summer Internship · Kyushu University, Japan
★ Results Published · ICHEP 2024
Transformer Feature Engineering Transfer Learning
Learn more →
Jet tagging at the LHC: collision to jet reconstruction to tagging

From collisions to discovery

A discovery in high energy physics is a statistics problem before it is anything else. The particle decay event we care about may appear in barely one out of several billion. To find it, we reconstruct the particle tracks, carefully select the signal-like events, and fit the kinematic variables that should highlight a signal.

Nowadays, the hand-tuned selection cuts are reaching their limit in modern collider experiments. The unprecedentedly complex detector designs produce data that is too high-dimensional, therefore deep learning neural networks (e.g. transformers) now correlate the subtle connections across dozens of variables at once. The challenges are real: models are trained on imperfect simulation, and yet to be interpretable enough that a collaboration will trust a result built on them. Getting that trade-off right is one of the defining problems of physics analysis today, and the part I most want to work on.

Global by design

I have lived and worked across five countries on three continents: China, Australia, Japan, the United Kingdom, and now Switzerland. Conducting internships and research projects within diverse international collaborations across different nations trained me to navigate language and cultural barriers while working effectively under high-demand conditions.

Mandarin Chinese
Mother Tongue
English
Native-level
Japanese
Fluent · JLPT N1
French
Intermediate · CEFR B1

Beyond the lab

2025 – 2026

Ski & Snowboard Instructor

École Suisse de Ski, Villars, Switzerland →
  • Delivered 182.5 hours of private ski and snowboard lessons to English- and Mandarin-speaking clients over the 25–26 winter season.
  • Tailored lesson content dynamically based on clients' age, fitness, and technical proficiency to ensure safe and high-quality skill development.
  • Supported resort operations by managing guest relations, providing navigation assistance, and administering emergency help during on-slope accidents.
Ski & Snowboard Instructor 1 Ski & Snowboard Instructor 2 Ski & Snowboard Instructor 3 Ski & Snowboard Instructor 4
2025 – 2026

Student Mountain Guide

Crazy Hikers Hiking Club, Switzerland →
  • Planned hiking routes across challenging Swiss alpine terrain, rigorously assessed real-time weather data, topographical maps, and terrain hazards.
  • Educated novice hikers on technical equipment selection, weather forecasting, and mountain navigation to ensure autonomous and safe route execution.
  • Co-led alpine excursions for groups of 10 to 30 participants, dynamically adapted route decisions based on group capabilities.
Student Mountain Guide 1 Student Mountain Guide 2 Student Mountain Guide 3 Student Mountain Guide 4
2023 – 2024

President · Kendo Club

Imperial College Kendo Club, UK →
  • Represented the club by liaising with the Imperial College Union, actively collaborating with the Management Board Group on strategic initiatives.
  • Chaired weekly committee meetings to drive alignment, streamline internal communications, and facilitate seamless cross-functional collaboration.
  • Managed operations and logistics for weekly training sessions serving 30+ participants, while overseeing budget and travel arrangements for national tournaments.
Kendo Club President 1 Kendo Club President 2 Kendo Club President 3 Kendo Club President 4

And many more ...

And many more 1 And many more 2 And many more 3 And many more 4

Built to work with people, not just problems

Physics teaches me to sit with hard problems alone, but the experiences above — coaching on a slope, leading on a mountain, coordinating a club — trained something different: the ability to read a room, adapt a message, and move with a group rather than ahead of it.

Clarity under pressure. Trust earned through accountability. The habit of explaining things until they actually land. That is what I bring to a team: not just analytical rigour, but the interpersonal range to boost the efficiency of the entire group. I have learned that the best ideas go nowhere without communication, and the best teams are built on more than shared technical objective.