I am a Senior Applied Scientist with Amazon in Seattle, US. I am currently working on agentic frameworks that orchestrate generative models for end-to-end video ad creation through natural language interaction. Previously, I worked on large-scale training of generative models for text-to-image generation. I did my PhD at University of Barcelona, where I was advised by Sergio Escalera, focusing on deep learning for facial action unit recognition. During my PhD I was a Visiting Scholar in Aleix Martinez’s lab at the Ohio State University, where I used algebraic topology to study how deep networks learn.
Contributing to Amazon Ads' agentic AI creative tool — an end-to-end system that produces professional video ads through conversational AI. Built the video transformation pipeline enabling automated multi-format creative production at scale.
Amazon Stores — ASIN-Centric Data at Scale for Subject-Driven Generation (2025)
Built an automated pipeline to curate millions of paired product image samples from the Amazon Catalog for training subject-driven generation models.
Amazon Stores — QA for Synthetic Images (2024)
Developed automated quality assessment of realism for AI-generated product images. Research accepted at CVPR 2025.
Trained large-scale text-to-image generative models for product contextualization on Amazon's Product Detail Page. Generated lifestyle imagery that places products in relevant scenes, reducing dependency on professional photography at scale.
Amazon Stores — Deep Catalog (2022–2023)
Created a 1.36 billion text/image dataset spanning 2,133 product types across 9 marketplaces. Curated and structured the Amazon product image catalog enabling large-scale training of generative models for product contextualization and subject-driven generation.
Tawny — Emotion AI (2021)
Built the computer vision core for Tawny's Emotion AI platform — real-time facial expression analysis that gauges human emotions from camera feeds. Technology deployed across market research, customer service, and retail behavioral analytics.
Selected Publications
Be More Specific: Evaluating Object-centric Realism in Synthetic Images
Anqi Liang, Ciprian Corneanu, Qianli Feng, Giorgio Giannone, Aleix Martinez
Computer Vision and Pattern Recognition, CVPR (2025)
