Yilin (Larry) Li (李亦林)
Hi! I'm Larry, a Data Scientist at Arteria AI and a Mechatronics Engineering graduate from the University of Waterloo with an AI Option. Previously, I worked as a Data Scientist at Telus and as a Machine Learning Engineer Intern at Huawei Canada.
Technical Focus
My work focuses on document extraction systems, multi-agent workflows, and production ML pipelines. I’m particularly interested in:
- Vision Language Models (VLMs) for document understanding
- Information retrieval and semantic search
- Applied reinforcement learning for decision-making systems
Fun fact: I got the CCOM performance certificate of Erhu (a traditional Chinese instrument) in grade 10!
Skills
Languages
Python, C++, C#, SQL, Java, Matlab, R
Frameworks
PyTorch, TensorFlow, Hugging Face (Transformers), CrewAI, Scikit-learn
Cloud & MLOps
GCP (Vertex AI, BigQuery), AWS (Textract), Docker, Databricks, ONNX Runtime
Work Experience
Data Scientist
Arteria AI • Toronto, ON
Building production document extraction systems for financial institutions using AWS Textract, Vision Language Models (Qwen, Idefics), and multi-agent workflows with CrewAI.
Data Scientist
Telus • Toronto, ON
Designed Offline RL agents for HVAC optimization (20% energy cost reduction) and production ML pipelines on GCP Vertex AI. Led fiber network fault detection project reducing MTTR by 40%.
Machine Learning Engineer Intern
Huawei Canada • Montreal, QC
Implemented 8-bit Quantization Aware Training on BERT encoders, compressing model size by 75% while retaining 98% accuracy. Applied knowledge distillation and structured pruning.
Deep Learning Engineering Intern
Synapse Technology • Palo Alto, CA
Developed 2D and 3D object detection algorithms for weapon detection in security applications.
Machine Learning Developer Intern
Primate Labs • Toronto, ON
Developed deep learning applications for Android device benchmarking and performance testing.
Robotics Software Developer Intern
Engineering Services Inc. • Toronto, ON
Developed navigation algorithms for autonomous indoor robots.
Research Experience
Undergraduate Research Assistant
University of Waterloo (Data Systems Group) • Advisor: Prof. Jimmy Lin
Co-developed a multi-stage retrieval system using Neural Query Synthesis with T5-3B. Achieved 1st place at TREC 2021 (0.71 nDCG) by solving the cross-encoding bottleneck for long document pairs.
Undergraduate Research Assistant
University of Waterloo (Kimia Lab) • Advisor: Prof. Hamid Tizhoosh
Researched medical image search and keyword extraction for pathology report analysis.
Projects
Smart Kitchen Multi-Agent System (SKMS)
Python, Google ADK, Gemini, Arize Phoenix
Built a hierarchical agent system with a stateful "Soft-Lock" protocol for concurrency control. Features 98% unit conversion accuracy and LLM trace observability via Arize Phoenix.
Education
University of Waterloo
B.A.Sc. in Mechatronics Engineering with AI Option
GPA: 90.6/100 • Dean's List