• Led the design, development, and maintenance of a Python package that hosts a suite of Kubeflow components used to construct flexible Kubeflow Pipelines for training XGBoost based ML models. This parallelized and streamlined the model training process for our data science partners.

• Engineered and managed infrastructure for the deployment of several ML models to cloud based production environments. The underlying infrastructure for the deployment is built on Python, Helm, Kubernetes, Seldon and Jenkins.

• Designed and conducted extensive load tests for APIs using Apache Jmeter which helped to boost overall API efficiency and performance.

• Designed and implemented new features on and maintained the team’s daily batch data transformation and publishing pipelines. Built on Java and hosted on AWS using AWS EMR, Step Functions, Lambda, EC2 and S3.

• Spearheaded several disaster recovery, failover and regional isolation exercises which greatly improved the security and resiliency of the team’s pipelines and workflows.

• Enabled enhanced logging, monitoring and dashboarding through technologies like AWS CloudWatch, Splunk, New Relic and PagerDuty.

• Designed, developed, and owned new features on a Python Model Monitoring package that generates Plotly visualizations, which enabled detailed analysis into ML model accuracy and stability.

• Leveraged data engineering packages like Pandas, Polars, DBT and Dask, along with database and storage technologies like Snowflake and Amazon S3.

• Facilitated CICD by creating end to end pipelines for various projects using Apache Airflow DAGs and Jenkins.

• Developed and redesigned an ATDD (Acceptance Test Driven Development) framework using Cucumber, Gherkin, Mimeo (An in-house API virtualization software) and Java.

• Improved readability and implemented API virtualization in the backend to enhance manipulation of testcase input parameters.

• Developed and implemented a web application which acts as a drone inventory allowing the user to read and manipulate inventory data.

• Used a MySQL database to store and synchronize all the drone data from the drone inventory.

• Hosted application using Heroku and implemented it using Node.js, React.js, HTML, and CSS.