• Leading and designing revenue growth projection project through complex data understanding and wrangling in Python, enhancing the accuracy of base rates in company valuations using multivariate regression models.• Building ETL pipeline to extract financial time series data from Refinitiv and loading the cleaned data into AWS RDS after transforming the data into non-time-series features.• Researching ways to implement document understanding language models such as LayoutLMv2 for extracting key financial indicators from company earning reports.

• Implemented Deep Convolutional Generative Adversarial Networks (DCGANs) using Tensorflow Keras library to generate photos of unique anime characters.• Coded and modified Residual U-blocks with different depths in PyTorch to efficiently implement U-squared Network for inferencing AI-generated human portraits.• Built, compiled, and trained the CycleGAN using Tensorflow to convert landscape photos into Monet-esque paintings by enforcing Cycle Consistency Loss.• Deployed containerized AI applications by creating docker images from repositories containing model architectures and trained weights.

• Contributed to the design of a novel method for group-summarised adversarial attack on vision models such as ResNets.

- Built a Machine Learning project in Python to predict auto insurance policy churn for a 12-month performance window with a curated model design and target definition.- Prepared >10M rows of train/validation/test data in Microsoft SQL by understanding and cleaning raw data with >500 features on the driver, policy, and VIN level.- Performed feature transformation and dimensionality reduction using statistical analysis, wrapper methods, embedded methods, and PCA, reduced # of features from >500 to <20.- Hyper-parameter tuned and cross-validated Logistic Regression, tree-based ensemble models, SVM and MLP model using grid search, achieved a maximum model lift of 412.- Conducted weekly and monthly financial forecasts, analyzed Prod and POV data to support financial activities (annual budget, data product subscription model, etc.) using Excel and MicroStrategy.- Reduced 10 hrs of forecast preparation time per month and 2% forecast error with VBA functions.

- Held independently virtual live labs for Stats 2B03 (Statistical Methods for Science), answering student questions and providing guidance on the use of R for the course.- Helped facilitate and aid in the Math Help Centre, resulting in effective assistance and communications with first-year Math & Stats students.

- Identified key data sources on Canadian real estate from regional real estate boards, CMHC, etc., assisting in building a robust data bank to support activities in the economic development office.- Analyzed residential and non-residential markets using reports from various sources (CBRE, JLL, Colliers, etc.); Used Excel and Python to analyze data and understand Hamilton’s key strengths in attracting local investment.- Developed a comprehensive user document and data catalogue for different aspects of Hamilton local economy.