- Led & Built Eclipse uProtocol’s socket communication which sends/receives interoperable protobuf messages between apps, and mocked test apps using the protocol to handle & respond to messages with synchronization and Behave test dev, reducing test runtime by 89% (open source: github.com/eclipse-uprotocol/up-tck)- Led & Built a friendship chemistry survey for data collection and a clustering AI matching algorithm, so new-graduate employees can seamlessly make new friends, resulting in ~80% increase in participants- Translated & routed simultaneous API messages to/from Event Hubs & Sockets using a multithreaded mediator in Java, allowing topic subscriptions and notifications between cloud users and the simulator- Created and simplified testing Java functions with JUnit Tests by isolating the testing classes’ behaviors via mocking the testing class’s dependencies using Mockito - Architected time-efficient database to manage users’ subscriptions & publications and implemented a scalable design to add/order message handlers in Python- Presented and taught to senior developers on AI basics and different models’ practicalities- Mentored and guided coworkers on steps to learn and employ software development best practices

▪ Customized PPO and GCN to lower network latencies by training each other to select optimal neighbors and node embeddings, respectively, using PyTorch▪ Implemented Adaptive Discretization Multi-Armed Bandit to “zoom” into the best GCN node embeddings, utilized the GCN’s output embeddings for the PPO’s environment, and customized PPO to converge to optimal node connections▪ Established network template to easily build/update graphs, calculate the shortest distance, and retrieve graph info ▪ Created network topologies in Python based on nodes’ peers, different topics, and PubSub roles▪ Presented the P2P network’s high latency problem, my algorithm’s implementation, and experiment analysis to senior committee members to inform them of the AI algorithm’s potential and future research directions

▪ Launched commercial UIs with integrated backend for admins to analyze & allocate resources▪ Designed codebase with TypeScript, CSS, and React design patterns to build scalable, clean frontend code▪ Updated, reviewed, and merged team’s code seamlessly using Git & Gitlab▪ Tested frontend display and human interactions with over 80% coverage in unit tests using Jest▪ Collaborated with my mentor to develop edge/error cases and to improve frontend designs and algorithms

▪ Integrated 4 data pipelines using Spark SQL and Java that preprocessed raw to consistent data for AI models▪ Tested transformations accuracies and implemented step definitions using Cucumber and Java▪ Used Agile Project Management to divide, describe, and assign stories and tasks to teammates

- Deployed a commercial Python program that optimizes parameters to fit experimental data to simulated data- Implemented objected-oriented prog, design patterns, and SOLID principles to create scalable, clean code- Employed fast computational methods, decreasing runtime by 92%, with parallelism, Pandas, & NumPy- Designed Tkinter GUIs so users can easily input data, compute, and visualize experimental and simulated fits

- Enhanced webpage designs, edited user/group permissions & fixed broken website links w/ Drupal and WordPress- Document instructions on fixing website problems- Test and analyze upcoming websites while providing feedback

- Created mathematical functions from experimental data utilizing Excel and Python- Co-managed Circle of Willis Model simulations/experiments