Leading the MLOps platform team, a team that built and managed a Kubernetes platform which enables 30 teams and 200+ users in deploying machine learning models, pipelines, and dashboards.

- Tech lead of MLOps platform team, maintaining and improving a self-service Kubernetes platform that enables data scientists to build data and ML Products that can scale.- Streamlined deployment process for ML models and data pipelines, enabling 20+ teams to deploy over 200+ pipelines and 20+ web services, including- Designed robust Airflow data pipeline patterns- Managed AWS and Kubernetes infrastructure (IAM/RBAC, S3, EC2, EKS, etc.) using Terraform, Flux, and Helm- Built standard Github Actions and CircleCI Orb pipelines that is used by all our users for easier deployment.- Led workshops and training sessions for 100+ users on how to fully utilize the platform- Built Helm charts that enable easy deployment of Redis Sentinel clusters- Built team documentation library, allowing a team of 3 in reliably supporting ~120 users via self-service documentation- Developed on call and emergency operations game plan, resulting in 99.8% uptime with a team of 3- Developed python CLI tool that automates various boilerplate commands- Managed Prometheus alert routing- Developed Grafana dashboard templates Cluster Components:K8 in AWS EKS. With Airflow, Kubeflow, Grafana/Prometheus, Flux, CircleCI

- Built and deployed a predictive machine learning model in R that quantifies customer credit risk- Developed Python ensemble model that predicts customer churn- Developed multi-level time series forecasting models in R to forecast sales goals for all of Grainger Canada salesforce- Automated data and prediction pipelines in Airflow, AWS, and Kubernetes- Trained team members in DevOps skills for automating data science projects- Led team efforts in documenting projects and clean code

• Created an ML-based pricing model (logistic, GMM) in R with an agile team, improving past solution by 20% in accuracy. • Developed production-level code of model above, resulting in a ready-to-ship Python module• Led the development of a Flask/D3.js/IBM Cloud web app which serves as the user-facing platform for the model.• Deployed MVP in IBM cloud and demoed it to a conference, resulting in buy-in and adoption of solution.• Communicated technical results and business recommendations to stakeholders.• Improved code quality and wrote detailed documentation for smooth handoff to engineering team.

• Developed and deployed a time-series anomaly detection model MVP using Python, Docker then experimented model with real-time data, resulting in detection of fraud within hours and saving the company from fraudulent orders.• Wrote and optimized SQL queries that extracted data from the BigQuery data warehouse for various ad hoc analyses.• Built an ETL system to scrape and load data from third party APIs/site content into BigQuery, enabling the marketing team to conduct daily competitor analyses.• Visualized customer data in Tableau and Python, interpreting data for POC of customer enrichment project.

Implemented automated data visualization and extraction features to an education software that leverages machine learning in teaching computer science concepts to secondary education students.