---

Heyuan LIU

Researcher & AI Engineer @ ENS-PSL / Inria / CNRS

Working on Theoretical Machine Learning, Explainable AI (XAI).

Heyuan LIU

Current Status

Research Engineer at ANTIQUE team (ENS-PSL/INRIA/CNRS)

Worked on Simplification for Neural Networks on Optimization

Previous

Research Intern OVGU MathOpt Group (Prof. Sebastian Sager who also leads Optimization Group in MPI-Magdeburg) in 2025

Research Intern EPFL IPESE lab (Prof. François Maréchal) in 2024

Education

MSc&T AiViC Graduate(MX23), École Polytechnique

Experience

Inria Logo

Research Engineer

ANTIQUE Team @ ENS-PSL/Inria/CNRS
Dec 2025 - Present

Focusing on the Neural Specification for Formal methods in Explainable AI under Prof. Caterina Urban.

OVGU Logo

Research Intern

OVGU MathOpt Group
Apr 2025 - Dec 2025

Simplification on post-trained ReLU Neural Networks embedded in MILP Systems under Prof. Sebastian Sager.

IPESE Logo

Research Intern

EPFL IPESE Lab
Mar 2024 - Sep 2024

Applied clustering/dim-reduction on energy datasets. Integrated LLM decision-making assistants.

Roland Berger Logo

Chatbot Engineer Intern

Roland Berger
Feb 2024 - Apr 2024

Developed Chatbots for car sales using Coze and Dify. Specialized agents for Sales and After-Sales.

VW Logo

Software Quality Intern

Volkswagen-Mobvoi
Jun 2022 - Aug 2022

Improved quality system from v2.0 to 3.0. Solved 36 software quality issues in ID6 and Audi A6 projects.

Education

Polytechnique

École Polytechnique

MSc&T in Artificial Intelligence & Advanced Visual Computing

Sep 2023 - Oct 2025(Certificates in June 2026)

GPA: 3.75/4.0 | SEMG Scholarship | Erasmus Scholarship

EPFL

EPFL

Exchange Master Student in IPESE Lab

Mar 2024 - Sep 2024

Energy Systems Modeling & Optimization

MUST

Macau University of Science & Technology (MUST)

B.Sc. in Software Engineering

Sep 2019 - Jun 2023

Graduated with Honor Degree

Research Interests

Explainable AI & Optimization

Computer Vision & NLP

Machine Learning Theory

PINN & Constraints Learning

Publications

On the role of Artificial Intelligence in Feature oriented Multi-Criteria Decision Analysis

H. Liu, Y. Zhao, and F. Maréchal

ESCAPE 35 (2025) Poster Presentation

Selected Projects

3D Luggage
Computer Vision

3D Luggage Detection

IDEMIA & École Polytechnique

Reconstructing 3D geometry of luggage items using multi-view images from airport conveyor systems. Tackling robust feature matching and geometric reasoning.

Read Report
DDIM
GenAI

Generative Image Editing

Diffusion Models

Modifying source images using DDIM inversion and Null-Text inversion ensuring feature retention while incorporating target prompts.

Read Report
EcoSystem
Unity 3D

3D EcoSystem Simulation

Computer Graphics

Scene construction in Unity with dynamic creatures, crowd simulation, and interaction logic starting from a flat terrain.

Watch Video
IPESE
Data Science

Multi-criteria Decision Analysis

EPFL IPESE

Data-driven methods for AI decision-making in Energy Systems using dimensionality reduction, clustering, and shape analysis.

View GitHub
StarCraft
Game AI

Real-Time AI for StarCraft

Reinforcement Learning

Implemented real-time AI strategies based on BWAPI. Ranked 1st in Terran and 2nd overall in the course competition.

View GitHub
NLP
NLP

Extractive Summarization

Graph Neural Networks

Utilized GraphSAGE, GCN, and GAT models to enhance text summarization by analyzing conversation structures.

View GitHub

Paper Reading Notes

Personal

Languages: English, Chinese, Deutsch (Basic), French (Basic)

Interests: Basketball, Saxophone, Dragon boat