Greater Vancouver Metropolitan Area
I am currently a product manager at Microsoft, working on improving the Native ads experience for both the demand and supply side of the marketplace. This could include running experiments on Selection algos, ranking and pricing algos or click prediction algos.
• Designed and maintained distributed low-latency ML models and infrastructure systems such as feature store and transformation/inference pipelines that are used by various ML teams for 90% of predictions and recommendations. • Built and deployed an end-to-end recommendation system which serves personalized blog posts using a two-tower recommendation architecture. This reduces latency by 30% • Developed an enhanced information retrieval HTTP API system by incorporating semantic search, using sentence-transformer embeddings and MongoDB vector search. This resulted in a 40% improvement in the quality and relevance of search results. • Collaborated with machine learning engineers and product managers to translate nontechnical requests, define project requirements, constraints and execution plans leading to a project completion rate of 100%.
• Conducted cutting-edge research in fraud detection using deep learning, and NLP techniques; leading to publications in reputable conferences and journals. • Fraud Detection Web Application: Built a web application using FastAPI, Tensorflow, MLflow, docker and Google App engine that accurately detects fraudulent websites by utilizing NLP techniques (to analyse unstructured text data)and machine learning classification models to classify websites based on their third-party connections (MSc Project).