Looking for a job

Yi Wang

@yiwang1

Message

I build end-to-end ML and on-device LLM applications for nutrition and food intelligence.

Zimbabwe

Message

What I'm looking for

I'm looking to build and deploy end-to-end ML, LLM, and on-device AI products, especially in nutrition, food intelligence, computer vision, and practical user-facing applications.

I've built end-to-end ML and LLM systems at the University of Maryland, including a domain-specialized model pipeline trained on 225K curated instruction-following QA pairs for nutrient estimation, food classification, and ontology-aware entity linking. My evaluations improved nutrient-estimation accuracy to 97% and nutrient-classification macro F1 to 96%.

At IAFNS, I developed Food Additive Lens, an offline iOS application that runs Llama-3.2-3B-Instruct-4bit on-device with Apple's MLX framework and RAG, delivering 30 token/s streaming responses.

My research at the University of Connecticut spans computer vision, multimodal vision-language learning, and real-time mobile inference. I've compressed vision models by 85.9% while preserving production-level accuracy, and I've published 12 peer-reviewed papers in machine learning and deep learning.

Experience

Work history, roles, and key accomplishments

Current

ML Engineer

Current

University of Maryland

Jan 2026 - Present (7 months)

Designed and built an end-to-end LLM training and evaluation pipeline, fine-tuning a domain-specialized model and curated, cleaned, and validated 225K instruction-following QA pairs for nutrient estimation, food classification, and ontology-aware entity linking. Evaluated model performance against Llama 3 8B, Gemini 2.0, and prompting-based baselines across multiple downstream tasks, improving nut