Building agentic AI for developer experience

• Overhauled the team’s user activity reports generation and maintenance process, built and maintained backend pipeline and packages, identified and resolved 20+ tickets..

• Refactored the dedicated report generation spark jobs into a single overarching spark job to produce all reports, while the vendor requirements are represented in JSON configuration files as inputs.

• Identified and resolved 5+ production-breaking discrepancies.

• Condensed 7000 lines of code down to 1500 lines, reducing system complexity by over 75%, with the use of Scala, Spark, Python, and Airflow.

• Shortened team’s new report onboarding time from over 20 days down to just under 3 days.

• Increased job runtime and decreased cost by 20%, by migrating to Spark 3.4.0 and AWS EMR 6.12.

• Working under the guidance of Professor Xifeng Yan, contributed to developing a task-oriented chatbot to teach and help users with cooking or DIY.

• Using Spacy and Panda, constructed a machine learning algorithm to extract important information such as

food names, cooking methods, time, ingredients, etc. with a final accuracy of 96%.

• Analyzed over 141,000 recipes in a database containing 210,000 wikiHow pages using python and HTML.

• Wrote code in RASA format, simulating over 100 possible user interactions in YAML based on user feedback.

• Developed a web based computer vision algorithm to aid researchers in analyzing neural tissue images.

• Using C++ and OpenCV library, the algorithm takes microscopic image stacks of nervous tissues, labels the

neurons in real-time, and outputs a 3D reconstruction of their locations. Achieved over 95% coverage of neurons.

• Migrated the algorithm to web using Node.js, creating a npm package with bindings to the OpenCV C++ library.

• Helped researchers improve efficiency by 80% when analyzing neural patterns and densities.