Lucas Jin Email & Phone Number

Toronto, Ontario, Canada

Conduct tutorials, hold office hours, grade assignments/exams, and answer Piazza questions for 800+ students in courses including Software, Network, Hardware, OS, Database, Architecture, Machine Learning, and Computer Vision, etc.

Vancouver, British Columbia, Canada

I have been a teaching assistant since my second year during my undergraduate at UBC. I was a bachelor of science (Computer Science) student there at UBC, I've TAed Computer Network (CPSC 317), Applied Machine Learning (CPSC 330) & Computer Hardware, Operating Systems (CPSC 313) and Software Construction (CPSC 210) courses, and some of them I TAed for multiple times. My main duty includes:● Holding weekly in-class activities and daily discussion boards (Piazza) for 300+ undergraduates who are from various departments and specializations.● Lecturing optional content (e.g., experiments, labs, homework extensions...) on tutorials and holding office hours to solve the difficult problems that students encountered throughout their studies.● Grading Assignments, giving back marks and feedback/advice, invigilating midterms/finals, and dealing with grading appeals requests.

Mountain View, California, United States

● Collaborated with Googler(Resident AI Researcher) and 2 peers, improved a Large Language Model’s performance by 7%.● Attended 10+ virtual events and expanded professional network, leading to successful applications to three top-tier graduate programs.

Vancouver, British Columbia, Canada

• Developed a CNN and an LSTM network model, using MODIS satellite data from 2000 to 2022, to predict plant blooming onset, achieved 19% improvement in prediction accuracy over the National Phenology Network’s First Bloom Index.• Integrated MODIS land cover classification, PRISM climate data, and geographical properties datasets with TensorFlow and scikit-learn, resulting in a 13.7% increase in the model’s generalization ability across diverse ecological regions and plant species.• Implemented transfer learning, spatial convolution, and hyperparameter tuning with GridSearchCV and RandomizedSearchCV, reducing the prediction error margin by 37% and setting a new baseline for remote sensing-based phenological predictions.• Led a multidisciplinary team (meteorological physicists, biologists, etc.) to process large-scale satellite data with AWS SageMaker. Presented at U of Washinton DSSG 2022, my model’s 22.3% improvement over benchmarks attracted top researchers’ interest.

Toronto, Ontario, Canada

• Co-designed, built, and supported log keeper plugins for the RBC Wealth Management Trading System using Spring Boot and PCF(Pivotal Cloud Foundry), serving 500,000 high-net-worth RBC clients and achieving a 99.8% system uptime.• Deployed and configured the ELK stack (Elasticsearch, Logstash, Kibana) to enhance log management and search capabilities, significantly improved resource accessibility and operational efficiency, reducing incident resolution time by 6.63%.• Optimized performance using Maven (dependency management), UrbanCode Deploy(automated deployment), and Jenkins(continuous integration/delivery), achieved an 8% reduction in deployment time while streamlining CI/CD pipelines.

Beijing, China

• Established a CUDA-based training environment from scratch, reduced training cost by 55.8% and boosted accuracy by 12.7%.• Pruned and fine-tuned YOLOv5(target detection), improved prediction efficiency by 37%, while maintaining an mAP over 0.85• Developed automated preprocessing scripts to unify heterogeneous datasets seamlessly. Implemented a data augmentationalgorithm increasing training data diversity by 33.3%, significantly enhancing the model’s generalizability at a minimal cost.• Implemented rule-based object detection algorithms to annotate datasets, significantly reduced manual labelling costs by 89%.

Beijing, China

● Utilized R-CNN to train on pre-labelled datasets, improving model precision by 11.3%.● Monitored mean Average Precision (mAP) on TensorBoard, achieving a consistent 0.81+ mAP score.● Developed auto-preprocessing scripts for heterogeneous datasets, increasing data processing speed by 40%.● Generated annotated videos with labels and bounding boxes, reducing manual labelling time by 90%

Master Of Engineering - Meng, Computer Engineering

Bachelor Of Science - B.Sc., Computer Science

Activities and Societies: Tri-mentoring program mentor

Bachelor Of Engineering - Be, Telecommunication Engineering, 3.7 - Undergraduate Tier 1 Scholarships

Activities and Societies: Philharmonic Chamber Choir of BUPT (Tenor section), Symphony Orchestra of BUPT(Percussion section)

Science

Activities and Societies: ZGCZX Golden Sail Wind Band, ZGCZX TV Station