• Focused on detecting Deepfake videos using generative convolutional vision transformers (GenConViT)• Finetuned GenConViT Deepfake video detection model on custom dataset, improved the accuracy and performance• Conducted comprehensive research on face detection models, comparing performance on detection speed, accuracy• Collected and cleaned video datasets, focusing on synchronizing audio and visual lip movements to train SyncNet• Published a deepfake detection survey paper titled The Tug-of-War Between Deepfake Generation and Detection toICML workshop 2024

• Researched, reviewed, and analyzed six LLM application evaluation frameworks to identify gaps in Intuit’s GenOsevaluation service, leading to the integration of DeepEval for more comprehensive coverage on evaluation• Designed and developed a generic interface enabling plug-and-play integration of third-party LLM evaluation librariesinto Intuit’s evaluation pipeline, utilizing Python, Kubeflow, Docker, and Kubernetes• Developed an automated pipeline to generate user-friendly evaluation reports for non-NLP experts, structuringevaluation metrics into a three-level hierarchy, and designed aggregated scores to simplify performance assessment• Automated the report distribution process by implementing an email notification component in the pipeline

• Developed a full-stack e2e Test platform utilizing React.js (frontend) and Java Spring Boot (backend)• Designed and implemented a 9-page user interface utilizing Material-UI (MUI), facilitating the view, creation, update, and deletion of test plans, test suites, test cases, and test files• Engineered backend services by creating REST API to handle CRUD operations for the test plans and XML tax files • Leveraged AWS S3 to store the generated test files to enhance platform scalability and accessibility

• Worked with Prof. Zhang and Ph.D. student Ruotong Wang on a project to evaluate the current practices of recording information from online meetings and designing a meeting summarization artifact• Built a framework to categorize and evaluate the different mediums people used to keep track of information• Submitted paper Connected Conversations: Designing Information Artifacts that Bridge between Synchronous and Asynchronous Communication to CSCW

• Built a full-stack internal reward system with React.js and Java Spring Boot• Designed and implemented a reward dashboard UI and added GraphQL endpoints to provide scalable backendsolutions for reward data retrieval• Enabled automated PR config checks to condition the triggering of sandbox test environment for Turbotax Online• Experience with GitHub app development, installation, and accessed Github REST APIs by Golang• Wrote unit tests using JavaScript and Golang’s test and mock frameworks, including gomock, gock, testify, and Jest

• Focused on company’s database maintenance and updated existing tables in ODPS data lake• Conducted data validation and data cleaning with Java, SQL, and Apache POI; created a data catalog for all tables inthe data lake with java code and functional APIs• Used Alibaba Could’s DataWorks to automatically pull data from remote databases and complete daily updates• Participated in meeting service and logistics of council reelection and press conferences

• Worked in a computer vision project, aimed at launching a promotion for the after-sales experience for customers• Focused on using the LE Picture Management System to check the quality of photos from the technicians andmaintenance workers• Implemented a data tracking system with JDBC, and analyzed sales data from different retail stores via Tableau