• Designed and built a Python/ReactJS application that leverages Large Language Models (LLMs) to automatically assign ICD-10 and HCC codes to patient records, significantly enhancing medical coding accuracy and efficiency for healthcare providers

• Led the development of a customer data enrichment service in Java Spring, enabling bulk enrichment via BigQuery SQL (80% time reduction from previous model), individual profile retrieval through Elasticsearch, and delivery to customer SFTPs or CRMs

• Conceptualized and implemented Java Spring POC to generate personalized customer outreach emails, leveraging GPT-4 and customer behaviour model to automate key sales workflows and improve SDR efficiency by 20%

• Built ETL pipeline to periodically refresh billionaire wealth data from the Forbes API using Airflow and Python

• Automated customer wealth analytics report creation to provide data-driven insights on potential leads using Java, SQL and Vega, boosting conversion rate by 10%

• Investigated the ad ranking treatment to improve ICO (incremental conversion optimization), proving that the proposed treatment increased incremental conversion by 20%

• Architected tool to retarget ads on new audiences for true incremental conversion by filtering out customers with high purchasing intent, maximizing ROI of ad campaigns

• Designed and implemented web app to add, update and delete customer-targeting rules with React and GraphQL

• Trained COVID-Net CXR-2 on chest X-ray dataset, achieving sensitivity of 95.5% and positive predictive value of 97.0%

• Curated COVID-19 positive vs negative chest X-ray dataset with 19,203 X-rays and 16,656 patients from 51 countries, creating

the largest, most diverse, open-access COVID-19 X-ray dataset in the world

• Collaborated with clinicians and radiologists to validate COVID-Net CXR-2 on private datasets and tuned for generalizability

• Architected, developed and productionized serverless AWS Lambda service for streamlined diagnosis of sortation related failures at Amazon warehouses, eliminating (70 hrs of labour)/day through log analysis and classification

• Integrated diagnostic service with AWS resources (CloudWatch, S3, DynamoDB) to automate end-to-end flow including triggering, data handling, analysis, and report generation

• Implemented unit tests, integration tests, and live alarms / metrics dashboards to ensure functionality and robustness of service