My team built the data ingestion to process millions of targeted offerings for Rogers and Fido customers. Core components: Databricks to curate data from DataLake, Azure Data Factory for ETL, and various structured and non-structured storage systems.I also own a user-facing component that delivers the targeted offering to customers. It is supported by multiple micro-services at the back.

As part of my graduate studies I partnered with Coca-Cola Canada to help develop a scalable model architecture, using NLP, to extract and clean POS data. It was meant to be a human-in-the-loop system in which the model (group of T5 models) generates the output with a confidence score (not accuracy).

I used NLP (Natural Language Processing) before it was cool.Used NLTK and spaCY in Quickbooks Online to help extract tax fields from user input.Also worked on QuickBooks Online Workpapers and ProTax.

Take Microsoft's Dynamic 365 Suite of products, customize it, and give to clients.