cv

Doctor of Philosophy (Software Engineering + AI)

McGill University

Software Engineering

• Main Research Interests: Trustworthy and Reliable Artificial Intelligence, Verification for Deep Learning, Software Engineering for AI, Requirements Engineering, Model Generation, Model-driven Software Engineering, Large Language Models
• Combining software engineering methods and large language model for safe and reliable use of AI in domain-specific applications.
• Using software models to integrate domain-specific best practices into large language model applications.

Bachelor of Software Engineering

McGill University

Software Engineering

• CGPA: 3.99/4.00 (Dean's Honour List)

Research Associate Intern (part-time)

Huawei Canada

• Worked on a project for generating software models using large language models on model-based testing.
• Developed a self-consistency framework for large language models with graph outputs.
• Proposed a model generation method that significantly improved accuracy compared to the baseline.
• Contributed to two paper submissions currently under review at a top software engineering conference.

R&D Intern

Aggregate Intellect

• Researched methods for automated curriculum planning by predicting prerequisite relationships between concepts.
• Developed a link prediction model with sentence transformer and graph neural networks.
• Built different data process components to recognize and deduplicate concepts which become the prototype of a recommendation system in production.

Teaching Assistant

McGill University

• Head TA for undergraduate software engineering courses: Model-based software engineering (since Winter 2021), Requirement Engineering (since Winter 2021), and Software Validation (Winter 2022).
• Conducted online and in-person tutorials and provided office hours to assist students with assignments and projects.
• Effectively managed the course project within a GitHub organization for the entire class, implementing automated delivery of tutorials and projects to students' repositories.
• Implemented an auto-grading system for course projects continuing to be used in subsequent course offerings.

Machine Learning Engineer (academic collaboration)

Aggregate Intellect

• Participated in the research as part of the Mitacs Accelerate Program.
• Developing a production-ready recommendation system for educational contents.
• Utilizing large language models for automatic document generation with structured outputs.
• Establishing an automated deployment pipeline for machine learning models on cloud platforms.

Research Assistant

McGill University

• Proposed and implemented a new method for graph model generation to generate realistic graphs while consistent.
• Used linear regression to estimate the score of generated graphs with hill climbing to navigate the search space.
• Published the result in a top journal in model-based engineering.

Game Developer Intern

Behaviour Interactive

• Participated in the gameplay, testing, and UI development for a mobile game in Unity.
• Solved multiple platform-dependent legacy bugs and received positive feedback from the players.

Discussion Group Organizer

Graph Neural Network Discussion Group

Hosting a discussion group in a machine learning community with over 3000 members.

FRQNT B2X Doctoral Scholarship

Fonds de recherche du Québec

Mitacs Accelerate Grant

Mitacs

McGill Engineering Doctoral Award (MEDA)

McGill University

McGill Dean’s Honour List

McGill University

Summer Undergraduate Research in Engineering Award

McGill University

Robustness of Models for Code

• Researched the robustness of graph neural networks and large language models for program code
• Applied sound certification methods for code-specific changes on the input data.
• Analyzed factors impacting the robustness with different properties on source code such as type and data flow.

Constraint Optimization for Scene Graph Generation

• Analyzed various fault patterns for machine learning models in object recognition, scene graph generation and visual question answering (VQA) and found that domain-specific constraints can be violated in existing models.
• Decouples the scene graph generation into multiple steps to avoid expensive training of large models.
• Proposed a post-processing method based on constraint optimization to ensure that domain-specific constraints in scene generation are always captured..

Incremental Graph Isomorphism Checking for Design Space Exploration

• Participated in the research as a summer undergraduate research program.
• Proposed an incremental graph isomorphism checking algorithm that boosts checking performance and efficiency.
• Contributed to an open-source graph generator (VIATRA-Generator).