Christopher Wang Email & Phone Number

Ottawa, ON, CA

Ottawa, Ontario, CA

Ottawa, Ontario, CA

• Designed and implemented, in a small cross-functional tiger team, a retrieval-augmented generative AI backend system using the OpenAI API, FastAPI, Azure Cognitive Search, and MongoDB, with automatic CI/CD integration.
• Crafted a novel New Product Introduction (NPI) algorithm for demand forecasting using the PySpark DataFrame API, increasing the accuracy for new product forecasts by 35% (from 0.85 WMAPE to 0.50 WMAPE)

Ottawa, Ontario, CA

• Architected a data segmentation module using the Databricks Job API, PySpark, and Argo Workflows, enabling parallel execution of downstream modules like inference and training, allowing for the processing of.

Ottawa, Ontario, CA

• Integrated multiple machine learning modules into a single unified Argo Workflow for streamlined deployment
• Led end-to-end ownership of the forecast explainability feature based on SHapely Additive exPlanations (SHAP)

Ottawa, Ontario, CA

• Performed hierarchal remote sensing (a class of image segmentation) on environmental satellite imagery using unsupervised random forests and unsupervised convolutional neural networks in R and Python

Burlington, MA, US

• Implemented homomorphically encrypted representation search using Face-Net embeddings and Microsoft SEAL
• Enabled efficient encrypted dot product by exposing the sum_vector operation in the TenSEAL open source library
• Performed literature review on secure and private inference using machine learning models, with a focus on cryptographic solutions

Ottawa, Ontario, CA

• Applied a Resnet-18 (convolutional neural network) on orca whale call data from the INTERSPEECH 2019 Computational Paralinguistics (ComParE) Orca Activity Subchallenge and Fisheries and Oceans Canada using the same.

Ottawa, Ontario, CA

• Prototyped various architecture designs for an automatic time series forecasting using technologies like Apache Kafka, Spark, Dask, Pandas, LightGBM, PrestoDB, Amazon S3, and Postgresql/PostGIS on a Kubernetes Cluster
• Operationalized the time series AutoML architecture using cloud technologies like Docker, Kubernetes, and Helm, following SOLID principles, agile development, and Python best practices, bringing the prototype to.
• Wrote and deployed a Postgresql Helm chart with prometheus sidecar for logging and metrics

Ottawa, Ontario, CA

• Experimented with neural networks (LSTMs, 1D convolutions), ensemble tree models, SVM regressors and feature selection techniques (genetic algorithms, mutual information, Boruta) for times series forecasting
• Introduced Dask on Kubernetes for parallelizing python jobs, resulting in linear speed ups (up to 16x)

Ottawa, Ontario, CA

• Achieved a competitive (state-of-the-art) mean segmentation accuracy of 92.2% on the research test set
• Authored "U-Net inspired architecture ensembles for left atrial segmentation" for SPIE Medical Imaging Conference 2019

Ottawa, Ontario, CA

• Designed a 3D biomedical image segmentation deep learning ensemble that leverages state-of-the-art feature extractors such as inception modules, residual connections, dense blocks, squeeze/excitation blocks, attention
• Developed a framework for rapid prototype testing of image segmentation models with data augmentation (translation, gaussian blur, elastic transform) for 2D and 3D images built on Keras/TensorFlow

Master Of Science - Ms, Computational Statistics And Machine Learning

Bachelor Of Engineering - Beng, Software Engineering

Ontario Secondary School Diploma