About

I am a Senior Applied Scientist at AWS AI Labs, where I develop research that powers the next generation of multimodal foundation models. Check out a couple recent examples here and here. Earlier this year, we launched Titan Image Generator, Amazon’s first text-to-image foundation model! Before this, I developed the NLP that powers Amazon’s physical store for clothing, Amazon Style. I received my PhD from Stanford University under the supervision of Johan Ugander, where I focused on designing and analyzing algorithms for preference learning from human feedback. My work commonly falls at the intersection of Natural Language Processing, Computer Vision, and Preference Learning.

I interned at Facebook CDS in the Summer of 2020, and helped develop ways to use People You May Know (PYMK) to catch spammers. In the Summer of 2019, I interned at Stitch Fix where I helped develop numerous approaches to better understand Client and Stylist preferences. Before my graduate work I received a BS in Electrical Engineering, Mathematics, and Economics from the University of Wisconsin, where I was advised by Barry Van Veen and Robert Blick. I have also interned at a number of companies in the past, including Microsoft and Silicon Labs.

Preprints

• L Zancato, A Seshadri, Y Dukler, A Golatkar, Y Shen, B Bowman, M Trager, A Achille, S Soatto B’MOJO: Hybrid State Space Realizations of Foundation Models with Eidetic and Fading Memory arXiv, July 2024.
• D Liu, A Seshadri, T Eliassi-Rad, J Ugander Re-visiting Skip-Gram Negative Sampling: Dimension Regularization for More Efficient Dissimilarity Preservation in Graph Embeddings arXiv, April 2024.
• A Seshadri, J Ugander Fundamental Limits of Testing the Independence of Irrelevant Alternatives in Discrete Choice (Journal Version) arXiv, January 2019.

Publications

• B Biggs^*, A Seshadri^*, Y Zou, A Jain, A Golatkar, Y Xie, A Achille, A Swaminathan, S Soatto Diffusion Soup: Model Merging for Text-to-Image Diffusion Models, To Appear in the European Conference on Computer Vision (ECCV), 2024.
• A Xu, M Vasileva, A Dave, A Seshadri HandsOff: Labeled Dataset Generation With No Additional Human Annotations Conference on Computer Vision and Pattern Recognition (CVPR), 2023. CVPR Highlight Award (Top 2% of Submissions)
• A Awadelkarim, A Seshadri, I Lo, I Ashlagi, J Ugander Rank-heterogeneous preference models for school choice Knowledge Discovery and Data Mining (KDD), 2023.
• S Verma, A Beniwal, N Sadagopan A Seshadri (2022) RecXplainer: Post-hoc Attribute Based Explainability for Recommender Systems, NeurIPS Trustworthy Embodied AI Workshop, 2022. Best Student Paper Award
• K Sevegnani, A Seshadri, T Wang, A Beniwal, J McAuley, A Lu, G Medioni Contrastive Learning for Interactive Recommendation in Fashion SIGIR Workshop On eCommerce, 2022.
• A Seshadri, S Ragain, J Ugander Learning Rich Rankings Neural Information Processing Systems (NeurIPS), 2020.
• A Seshadri, J Ugander Fundamental Limits of Testing the Independence of Irrelevant Alternatives in Discrete Choice ACM Conference on Economics and Computation (EC), 2019.
• A Seshadri, A Peysakhovich, J Ugander Discovering Context Effects from Raw Choice Data International Conference on Machine Learning (ICML), 2019.
• A Bhat, PV Gwozdz, A Seshadri, M Hoeft, RH Blick Tank Circuit for Ultrafast Single-Particle Detection in Micropores Physical Review Letters, 2018
• E Stava, HC Shin, M Yu, A Bhat, P Resto, A Seshadri, JC Williams, RH Blick Ultra-stable glass microcraters for on-chip patch clamping