Viraj Prabhu

Viraj Prabhu

I am a research scientist on the multimodal AI team at Salesforce Research. I received my PhD in Computer Science from Georgia Tech in December 2023, where I was advised by Judy Hoffman and worked on making computer vision models generalize to new environments. I earned my Master's in CS (awarded the MS Research award) in 2019, also at Georgia Tech, where I was advised by Devi Parikh and worked on developing visual conversational agents.

In grad school, I've had the opportunity to intern at NVIDIA (with Sanja Fidler), Salesforce (with Nikhil Naik), and Curai (with Anitha Kannan). Before that, I've had stints as a research assistant at Virginia Tech (with Dhruv Batra) and Adobe, and as a mentor for Google Summer of Code. I received my Bachelor's degree in Computer Science from BITS Pilani in 2015.

In my free time, I enjoy reading, running, soccer, and playing the guitar.

Email  /  CV  /  Scholar  /  LinkedIn  /  Github

BITS Pilani 2011-2015
Georgia Tech 2017-2023
Adobe Summer 2014
2015-2016
Curai Summer 2018, 2019
Salesforce Summer 2021
Jan 2024-current
NVIDIA Summer 2022

Research

My current research focus is on developing multimodal digital agents that can perceive, reason, and act in novel environments to accomplish complex goals. I'm particularly excited about LLM-modulo approaches that enhance and robustify LLMs using structured outputs and external verifiers. Representative publications are highlighted.

Trust but Verify: Programmatic VLM Evaluation in the Wild

Viraj Prabhu, Senthil Purushwalkam, An Yan, Caiming Xiong, Ran Xu
ICCV 2025
PROVE

xGen-MM (BLIP-3): A Family of Open Large Multimodal Models

Salesforce AI Research (Viraj Prabhu: core contributor)
ICCV 2025 Findings Workshop
BLIP-3

We're Not Using Videos Effectively: An Updated Video Domain Adaptation Baseline

Simar Kareer, Vivek Vijaykumar, Harsh Maheshwari, Judy Hoffman, Prithvijit Chattopadhyay, Viraj Prabhu
TMLR 2024
Video Domain Adaptation

Translating Labels to Solve Annotation Mismatches Across Object Detection Datasets

Yuan-Hong Liao, David Acuna, Rafid Mahmood, James Lucas, Viraj Prabhu, Sanja Fidler
ICLR 2024
Translating Labels

AUGCAL: Sim-to-Real Adaptation by Improving Uncertainty Calibration on Augmented Synthetic Images

Prithvijit Chattopadhyay, Bharat Goyal, Bogi Ecsedi, Viraj Prabhu, Judy Hoffman
ICLR 2024
AUGCAL

LANCE: Stress-testing Visual Models by Generating Language-guided Counterfactual Images

Viraj Prabhu, Sriram Yenamandra, Prithvijit Chattopadhyay, Judy Hoffman
NeurIPS 2023
LANCE

Bridging the Sim2Real gap with CARE: Supervised Detection Adaptation with Conditional Alignment and Reweighting

Viraj Prabhu, David Acuna, Yuan-Hong Liao, Rafid Mahmood, Marc T. Law, Judy Hoffman, Sanja Fidler, James Lucas
TMLR 2023
CARE

Battle of the Backbones: A Large-Scale Comparison of Pretrained Models across Vision Tasks

Micah Goldblum, Hossein Souri, Renkun Ni, Manli Shu, Viraj Prabhu, Gowthami Somepalli, Prithvijit Chattopadhyay, Adrien Bardes, Mark Ibrahim, Judy Hoffman, Rama Chellappa, Andrew Gordon Wilson, Tom Goldstein
NeurIPS Datasets & Benchmarks 2023
Battle of Backbones

FACTS: First Amplify Correlations and Then Slice to Discover Bias

Sriram Yenamandra, Pratik Ramesh, Viraj Prabhu, Judy Hoffman
ICCV 2023
FACTS

ICON2: Reliably Benchmarking Inequity in Detection by Identifying and Controlling for Confounders

Sruthi Sudhakar, Viraj Prabhu, Olga Russakovsky, Judy Hoffman
CVPR Workshop 2023
ICON2

Can domain adaptation make object recognition work for everyone?

Viraj Prabhu, Ramprasaath R. Selvaraju, Judy Hoffman, Nikhil Naik
CVPR Workshop 2022
Domain Adaptation for Everyone

Mitigating Bias in Visual Transformers via Targeted Alignment

Sruthi Sudhakar, Viraj Prabhu, Arvind Krishnakumar, Judy Hoffman
BMVC 2021
ViT Bias Mitigation

Adapting Self-Supervised Vision Transformers by Probing Attention-Conditioned Masking Consistency

Viraj Prabhu*, Sriram Yenamandra*, Aaditya Singh, Judy Hoffman (* = equal contribution)
NeurIPS 2022
PACMAC

AUGCO: Augmentation Consistency-guided Self-training for Source-free Domain Adaptive Segmentation

Viraj Prabhu*, Shivam Khare*, Deeksha Kartik, Judy Hoffman (* = equal contribution)
ECCV Workshop 2022
AUGCO

UDIS: Unsupervised Discovery of Bias in Deep Visual Recognition Models

Arvind Krishnakumar, Viraj Prabhu, Sruthi Sudhakar, Judy Hoffman
BMVC 2021
UDIS
SENTRY

Active Domain Adaptation via Clustering Uncertainty-weighted Embeddings

Viraj Prabhu, Arjun Chandrasekaran, Kate Saenko, Judy Hoffman
ICCV 2021
CLUE

Open Set Medical Diagnosis

Viraj Prabhu, Anitha Kannan, Geoffrey J. Tso, Namit Katariya, Manish Chablani, David Sontag, Xavier Amatriain
NeurIPS Workshop 2019
Open Set Medical

Few-shot Learning for Dermatological Disease Diagnosis

Viraj Prabhu, Anitha Kannan, Murali Ravuri, Manish Chablani, David Sontag, Xavier Amatriain
MLHC 2019 (spotlight)
Few-shot Dermatology

Do Explanations make VQA Models more Predictable to a Human?

Arjun Chandrasekaran*, Viraj Prabhu*, Deshraj Yadav*, Prithvijit Chattopadhyay*, Devi Parikh (* = equal contribution)
EMNLP 2018
VQA Predictability

The Promise of Premise: Harnessing Question Premises in Visual Question Answering

Aroma Mahendru*, Viraj Prabhu*, Akrit Mohapatra*, Dhruv Batra, Stefan Lee (* = equal contribution)
EMNLP 2017
Premise

Evaluating Visual Conversational Agents via Cooperative Human-AI Games

Prithvijit Chattopadhyay*, Deshraj Yadav*, Viraj Prabhu, Arjun Chandrasekaran, Abhishek Das, Stefan Lee, Dhruv Batra, Devi Parikh (* = equal contribution)
HCOMP 2017
Visual Conversational Agents

Miscellaneous

Service & Recognition

Talks & Media

Projects & Software