ML Engineer · AI Fairness · Montreal

Kelsey

王佳宜

Machine Learning Engineer & Consultant at Synechron. M.Sc. from McGill University (Mila).

01

About me

Originally from Xi'an, China — I moved to Montreal for university, completing a B.S. in Mathematics and an M.Sc. in Mathematics & Statistics at McGill, researching deep learning and AI fairness at the Oberman Lab.

I now build RAG pipelines, agentic AI workflows, and full-stack ML systems for financial services clients at Synechron.

RAG Pipelines AI Fairness Deep Learning Agentic AI Optimization PyTorch
02

Work experience

Sept 2024 — Mar 2026

Synechron

Montreal, Canada

Consultant & ML Engineer  ·  View projects ↗

  • Built reusable RAG pipelines and agentic AI workflows, standardizing ML patterns across financial services projects.
  • Architected an AI-powered call center training platform — cut ML inference from 15s to under 3s.
  • Contributed to RFP responses covering AI governance, regulatory compliance, and bias mitigation.

Jan 2024 — Aug 2024

D-BOX Technologies

Longueuil, Canada

Deep Learning Intern

  • Trained and fine-tuned multimodal models (PyTorch) for video and audio event detection.
  • Developed preprocessing pipelines for large-scale vision and audio datasets.

May 2022 — Dec 2022

Lambdanalytique

Montreal, Canada

Data Science Intern

  • Built an image preprocessing pipeline (Docker, CNN, ResNet) — 80% reduction in processing time.
  • Developed a genetic algorithm optimization pipeline for game development.
03

Research & publications

Sept 2021 — Jan 2024

Oberman Lab, McGill · Mila

Montreal, Canada

M.Sc. Researcher

  • Thesis on AI fairness — bias and equity in overparameterized deep learning models.
  • Reviewer for ICML and NeurIPS workshops on Algorithmic Fairness.
  • NSERC Master's Scholarship ($17,500) · MITACS Scholarship ($15,000) · Grad Excellence Award.

Adagrad Promotes Diffuse Solutions In Overparameterized Regimes

A. Rambidis, J. Wang  ·  OPT Workshop at NeurIPS 2023

Theoretical and empirical study showing that Adagrad promotes more diffuse solutions in over-parameterized settings, offering potential advantages for deep learning optimization.

04

Outside work

Tennis

On courts, in all weather.

Top Rope

Got into top rope recently!

Techno

Say no more.

Traveling

Next destination will be Albania:)

05

Get in touch

Always happy to connect — whether it's research, collaborations, or just a good conversation.