Chunfeng Li Email & Phone Number

Toronto, Ontario, Canada

Toronto, Ontario, Canada

• Creating tools and testplans to validate features and performance for ML models that are rewritten in C++ and Python. • Models in the modelzoo include LLMs such as GPT3, BERT and Llama 3

Waterloo, Ontario, Canada

• Created and deployed, using Jenkins, data ingestion pipelines on Azure DataFactory using the MDMF (Metadata Driven Management Framework) toprocess internal data for consumption automatically on daily intervals• Coordinated between several teams to oversee and manage the changemanagement process to ensure organised deployments to production• Used Pyspark to perform large-scale Parallelized data processing in adistributed environment to manipulate SQL data to conform to regulations

Toronto, Ontario, Canada

▪ Collaborated within the bank-wide Quartz Python workspace to addmultiple features to improve the functionality and usability of ourapplications that are used by traders within Bank of America▪ Implemented a multithreaded regression test harness for our application,effectively managing concurrent calls to a grid computer and ensuring theprevention of errors in the Sandra Global Object Database caused byconcurrency issues.▪ Utilized pandas to compile equity ticker data from multiple upstreamsources, enabling the provision of accurate data for applications

Toronto, Ontario, Canada

▪ Modified the existing AzureML pipeline and its scripts to detect newsymbol objects for clients and reduce memory usage using AMP▪ Created and designed a new AzureML pipeline that facilitates the fastretraining of the trained detectron2 object detection model to cut downtraining time for new client data by over 50%▪ Redesigned the existing pipeline to allow for modification of trainingconfigurations such as layer freezing and different training data weights▪ Using the new retraining workflow, increased previous detection 2 symboldetector model recall on general data by 2% and precision by 5%

Toronto, Ontario, Canada

▪ Trained and Tested TensorFlow mobile net object detection and yolov5models to improve on the existing Greeter AI app model on the App Store▪ Analyzed client data from deployed Greeter AI models to present metricsand insights to clients using Jupyter Notebook, Matplotlib and Pandas▪ Designed and created a multi-threaded Cloud inference server in Pythonthat communicates with the Greeter AI app and integrates with GoogleFirestore for storage of client metrics▪ Created and tested multiple different prediction models used to predictclient sales using Scikit-learn