I build streamlined Computer Vision Pipelines for early-stage startups to cut compute costs.

I have practical and academic research experiense atdistilling knowledge from foundation modelsto build smaller vision models that outperform foundation model with fraction of training infrastructure cost and fewer training data.   Built and deployed vision pipeline for 2 AI startups with 1000+ active users, saving$10K+ inference cost.I also have 1 year of experiense as a freelance computer vision engineer. Made (first-authored, academic research) a vision transformer interpretability method named F2HF, that is currently the state-of-the-art, widely adopted by industry and accepted in IEEE conference. With proven academic research and practical experience, I hold the ability to build and deploy ML based vision pipeline that are cost effective, and get the job done.