Research
Our research teams build self-correcting evaluation systems and e-commerce benchmarks for image and video models, so that generative AI delivers real business impact as it becomes part of how products are sold.
Self-correcting evaluation
Closed-loop systems where a model judges, repairs, re-verifies and learns from its own output, so quality improves without a human in every loop.
E-commerce benchmarks
Task-grounded benchmarks that measure whether generated product imagery is accurate, on-brand and sellable, not just whether it looks plausible.
Image and video harness
Studying how frontier image and video models fail on real catalog data, and what it takes to make them reliable for production storefronts.
Latest research
PublicationSep 7, 202628 min read
Building a Self-Improving Visual Evaluation Harness
How we built a closed-loop agent that can judge, repair, re-verify, and learn from failures in e-commerce image generation
Hrithwik BharadwajRead the post →
