• Built and own a distributed ML inference platform (0→1) serving ~25% of production scoring traffic, establishing a standardized execution layer for ML models.

• Designed a backend-agnostic inference abstraction (TensorFlow, NVIDIA Triton), decoupling model logic from runtime infrastructure and enabling flexible deployment.

• Migrated and standardized large-scale ML pipelines supporting revenue-critical workflows, improving reliability and consistency across scoring systems.

• Built scalable orchestration and data processing systems using Airflow (Python), Dataflow (Scala), and BigQuery to support feature computation and scoring at large scale.

• Reduced inference and infrastructure costs by optimizing execution paths and resource utilization (to <$3/day for initial workloads).

• Improved system reliability by replacing time-based scheduling with data-driven triggering, eliminating recurring production failures.

• Led incident response and root cause analysis for production latency issues, identifying data corruption and implementing structural fixes.

• Reduced CI/CD test runtime by 3–6× by redesigning execution into parallel, resource-aware batches.

• Drove adoption of platform infrastructure across team workflows, enabling consistent and reusable ML execution patterns.

• Developed a binary classifier to detect bullets in X-rays with a 90% precision score by conducting transfer learning using Python and TensorFlow

• Increased model reliability by increasing dataset size by over 38,225% by employing advanced data augmentation techniques

• Achieved high accuracy and robustness with an F1 score of 87% by utilizing model refinement and extensive dataset enhancement

• Leveraged Kotlin and Spring Boot to optimize the Entitlements API, resulting in a 50% decrease in error rates, a 20% increase in API response speed, and a 15ms reduction in loading time for critical web pages, significantly enhancing the user experience.

• Designed and developed API endpoints and a user-friendly frontend for a dynamic questionnaire management application, utilizing C# and .NET technologies. Achieved a substantial reduction in employee onboarding time, improved data security, and streamlined questionnaire updates, contributing to operational efficiency.

• Independently developed a testing framework using Python to validate the correctness of API endpoints

• Wrote production quality Swift and Python code to build an app called TreeSearch that uses HR information on over 1000 of WillowTree’s employees to create a database that can be filtered and searched by interacting with an intuitive UI

• Implemented all filtering and searching functionality of the IOS version of the app using Swift and SwiftUI

• Customized the backend response with Python and network libraries including Chalice

• Secured sensitive information in TreeSearch by collaborating with two senior software engineers to set up a Cognito user pool in AWS to allow for logging in and out of the app

• Assisted with integrating the Cognito user pool with G suite and Amplify to allow users to seamlessly login to TreeSearch using their Google account

"I was thoroughly impressed with Adam's skill as an engineer, especially in a discipline that he was not originally interviewed for at WillowTree. The stories Adam chose to work on all required him to learn a crash course in Python, Amazon Web Services, and DevOps practices. Adam worked hard and diligently on his tasks, and also strived to make sure he understood the solutions he was implementing. In addition, I was also impressed at how Adam took feedback with a positive attitude and his willingness to implement said feedback immediately in his day to day. Finally, I was encouraged to see that Adam was willing to work on a solution for a while himself, and then come to a discussion with pointed questions and explanations for issues he was encountering" - Clay Comer, Senior Software Engineer, WillowTree