Paloma Sodhi

paloma.sodhi at gmail dot com


I am a Research Scientist at ASAPP working on task-oriented dialog and natural language processing for conversational AI systems. I received my Ph.D. in Robotics at Carnegie Mellon University with Michael Kaess where my research was generously supported by the CMU Presidential Fellowship and Meta AI.

I am generally interested in the problem of predictive processing. My research looks at efficient machine learning techniques for inferring latent states online from observation sequences such as language, vision, and touch. My goal is to apply such techniques towards enabling seamless interactions between human and AI partners.

I am excited about product-driven research and have previously worked across applications like tactile manipulation, underwater exploration, and computational agriculture.


Mar, 2022 Successfully defended my Ph.D. on “Learning and Inference in Factor Graphs with Applications Tactile Perception”!
Jan, 2022 Our work, PatchGraph, on tracking unseen objects during in-hand manipulations accepted to ICRA 2022! Links: paper, code+video.
Jan, 2022 Gave a talk on factor graphs and tactile perception at the GRASP SFI seminar.
Dec, 2021 Check out my ML@CMU blog post on An Energy-based Perspective on Learning Observation Models.
Nov, 2021 Had a lot of fun co-designing a Technights session on Deconstructing robotics!
Nov, 2021 Excited to present our work on learning energy-based models in factor graph optimization at CoRL 2021! Links: paper, code+video.
Sep, 2021 Our work on learned GPR models selected as a best conference paper finalist at IROS 2021. Congratulations Alex! Links: arXiv.
Jul, 2021 Excited to be in the 2021 cohort of RSS Pioneers! Links: poster, statement.