Aarush Gupta
I am a ML Engineer with the Generative ML team at Snap Inc. working on efficient image and video foundation models.
I graduated from the MS in Robotics program at Carnegie Mellon University, where I worked with Laszlo Jeni, Jian Ren & Sergey Tulyakov from Snap Inc., on real-time neural rendering. Before graduate school, I worked on developing talking-head video generation models at Rephrase.ai.
Apart from work, I am enthusiastic about financial markets, guitar, food, and movies.
March 2025: SnapGen has been accepted to CVPR 2025 and also covered in TechCrunch!
September 2023: LightSpeed has been accepted to NeurIPS 2023.
September 2023: Joined the Generative ML team at Snap Inc. NYC.
August 2023: Finished master's program at Carnegie Mellon University.
May 2023: "Structure-Preserving Instance Segmentation via Skeleton-Aware Distance Transform" accepted at MICCAI 2023.
May 2022: Working with DIMAAG.AI on autonomous agricultural vehicles for the summer.
August 2021: Started graduate school at Carnegie Mellon University.
Preventing Shortcuts in Adapter Training via Providing the Shortcuts
Anujraaj Goyal, Gordon Qian, Huseyin Coskun, Aarush Gupta, Himmy Tam, Daniil Ostashev, Ju Hu, Dhritiman Sagar, Sergey Tulyakov, Kfir Aberman, Jackson Wang
NeurIPS 2025
Shortcut Rerouting re-enables text control of pose and expression after adapter training. Without shortcut rerouting, the adapter overfits to the reference image and reproduces its pose and expression, ignoring the prompt. With Shortcut-Rerouted Adapter Training, the adapter disentangles identity from other factors, allowing the model to respond faithfully to prompt-specified expressions and head poses.
SnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training
Jierun Chen*, Dongting Hu*, Xijie Huang,* Huseyin Coskun, Arpit Sahni, Aarush Gupta, Anujraaj Goyal, Dishani Lahiri, Rajesh Singh, Yerlan Idelbayev, Junli Cao, Yanyu Li, Kwang-Ting Cheng, Mingming Gong, S.-H. Gary Chan, Sergey Tulyakov, Anil Kag, Yanwu Xu, Jian Ren
CVPR 2025
SnapGen is the first image generation model (379M) that can synthesize high-resolution images (1024x1024) on mobile devices in 1.4s, and achieve 0.66 on GenEval metric.
LightSpeed: Light and Fast Neural Light Fields on Mobile Devices
Aarush Gupta, Junli Cao*, Chaoyang Wang*, Ju Hu, Sergey Tulyakov, Jian Ren, Laszlo Jeni
NeurIPS 2023
Master's Thesis Project
Structure-Preserving Instance Segmentation via Skeleton-Aware Distance Transform
Zudi Lin, Donglai Wei, Aarush Gupta, Xingyu Liu, Deqing Sun, Hanspeter Pfister
MICCAI 2023
MitoEM Dataset: Large-Scale 3D Mitochondria Instance Segmentation from EM Images
Donglai Wei, Zudi Lin, Daniel Franco-Barranco, Nils Wendt, Xingyu Liu, Wenjie Yin, Xin Huang, Aarush Gupta, Won-Dong Jang, Xueying Wang, Ignacio Arganda-Carreras, Jeff W. Lichtman, Hanspeter Pfister
MICCAI 2020
Reviewer service: AAAI (2026), CVPR (2025-26), NeurIPS (2025-26), TPAMI (2025), ICCV/ECCV (2024-25), FG (2024-25)
Led the Vision and Language Group at IIT Roorkee (2018-2020)
Teaching assistant, Faculty Development Program, IIT Roorkee (April 2018)
Vice-chair of ACM student chapter at IIT Roorkee (2019-2020)
Student Mentorship Program, IIT Roorkee (2018-2020)
SN Bose Scholarship (2019)
KVPY Fellowship (2015)
