Experience

Associate Applied Machine Learning Specialist

Builds foundation-model evaluation systems, benchmark-generation workflows, multimodal extraction pipelines, and efficient GPU inference. Recent work includes Fine-Grained Benchmark Generation and Open-PMC-18M.

Reduced Qwen2.5-VL extraction latency from 4.0 seconds to 1.2 seconds per sample and multimodal summary generation from 4.0 seconds to 0.6 seconds, while supporting OpenCLIP training on 18M image-text pairs.

Graduate Research Assistant

Built reusable multimodal ML pipelines for dataset construction, experimentation, evaluation, and error analysis. Developed agentic workflows and benchmarks for data storytelling, chart reasoning, and interactive dashboard question answering.

Selected work includes RL-Text2Vis, DashboardQA, ChartQAPro, and DataNarrative.

Lecturer

Designed and delivered theory and lab courses in programming, software engineering, and systems analysis while mentoring students in debugging, reliable software construction, and technical communication.