Led the development and integration of end-to-end ML microservices, deploying production-ready models that drove Grainger Canada's web search using name entity recognition and text classification algorithms

    Designed an LLM-based system to generate descriptions of website products, fine-tuning on existing high-quality descriptions, incorporating manual review for quality and improvement

    Refactored the category prediction fastText model by migrating it from an on-premises server to the cloud through MLflow and Databricks, integrating version control and ensuring reproducibility, leading to a 37% reduction in model training time and an 8% enhancement in recall

    Deployed an ML API that utilizes a GPT-3 powered model to assess parsed search queries from an established NER model to identify incorrect query labels, resulting in a notable reduction in the time taken

    Mentored a team of two junior data scientists and successfully reduced tech-debt of existing ML pipelines by 80%

    Executed health check scripts and stress testing for REST APIs using Locust, implemented efficient logging with Splunk, and utilized Grafana dashboards to monitor model drift and performance

      Fine-tuned BERT for classifying sentiment of 20,000 user reviews, integrating the result in the internal CRM tool, thereby improving the support teams response time to negative feedback

      Implemented clustering and segmentation for over 10 million investors and 37,000 brokers using K-Means and DBSCAN, for sending targeted notifications on app, increasing app usage by 25%

      Managed a team of two data analysts in writing efficient SQL queries and reducing time spent on building reports

      Created an item-item collaborative filtering model to recommend products to customers in the broker platform for upselling and cross-selling schemes, leading to an 11% increase in digital sales

      Built Power BI visualization reports with data from SQL Server & Google Analytics within an ETL framework