Kelvin Zhao

UC Berkeley MEng Capstone

Matterport Team • Removed recurring manual process by building a Temporal workflow into Matterport’s Kotlin GraphQL server; utilized protobufs, custom DSLs, AWS S3, and delivered with full unit and integration test coverage. • Developed and deployed a new microservice exposing previously inaccessible vision pipeline; containerized with Docker, configured Helm for Kubernetes, and automated CI/CD via Jenkins, Bash, and Makefiles. • Migrated and enhanced a company-wide Python CLI tool; added concurrent file download, data querying, and credential-resolution features, ensuring smooth rollout via stakeholder coordination and user feedback.

• Enabled developers to easily execute load tests on AWS EC2 instances and Kubernetes on-premise servers by implementing an internal load testing service with Next.js, Node.js, and TypeScript connected to Gatling and K6 microservices • Implemented live log streaming of remote load tests via WebSockets & artifact storage with DynamoDB/S3 • Led multiple POCs into practical enterprise solutions integrated into Paciolan’s internal tooling and test pipelines • Saved dev hours by improving reliability of testing infrastructure, improving automated Nightwatch report generation and integrating AWS services like DynamoDB with Grafana and SQS/Lambda with Slack apps

• Designed a new dynamic computing architecture for Texera (web-based no-code data analytics platform), leveraging Kubernetes for isolated workflow executions, enhancing scalability and flexibility under heavy data workloads • Developed a custom controller with Dropwizard, Kubernetes Java API, MySQL, and Envoy Proxy, exposing a REST API for K8s interaction, while managing workflow-specific pods for optimized execution/user experience • Integrated Docker and Helm to streamline Texera's stack deployment, simplifying local and production workflows

• Headed lab-wide shift from proprietary to open-source software, streamlining lab research and workflow efficiency • Constructed data pipelines which enhanced efficiency of lab's machine learning training process with Julia • Organized and loaded proprietary models for 3D medical imaging segmentation • Created and managed Python/Julia plugins in open-source software (3D Slicer), automating clinical processes and earning authorship in multiple research papers • Optimized legacy Matlab codebase using Julia to enable concurrent usage across multiple clinical facilities

Coursework: - Object Oriented Programming - Data Structures and Algorithms - Databases and Data Management - Parallel Computing and Distributed Systems - Computer Networks - Operating Systems - Machine Learning and Data Mining