Experience
Machine Learning Engineer Intern
I worked on AI Reviewer and Writing Feedback, building models, evaluation pipelines, and the services that support automated writing assessment.
Built the end-to-end evaluation and data pipeline for AI Reviewer, including metric design, iterative dataset curation and annotation, and scalable third-party labeling workflows.
Led development of the transformer-based reviewer in PyTorch. On IELTS scoring, the model achieved 88% accuracy within ±1.0 band and 97% within ±1.5 bands, approaching human-grader agreement on standardized scoring.
Migrated the writing feedback system from Prompt Flow and Flask to LlamaIndex and Quart. Redesigned API routing and introduced asynchronous request handling to improve scalability and reduce processing time on internal API endpoints by 30%.
The AI Reviewer work led to a pending U.S. patent on calibrating LLM review through human alignment.