PhD Candidate · AI Researcher
Hello—Namaste! I build multimodal and generative AI systems that can understand, create, and act reliably in the physical world.
New York City, from the water
one of my favorite perspectives
About me
I am a Computer Science PhD candidate at the University of Illinois Urbana-Champaign, advised by Dr. Ismini Lourentzou in the Perception and Language Lab. I previously completed my MS and BS in Computer Science at Virginia Tech.
My work spans multimodal learning, controllable image and video generation, embodied agents, uncertainty estimation, and efficient adaptation of foundation models. The question that keeps pulling me forward is simple: how can we make capable AI systems more understandable, dependable, and useful in the open world?
I try to bring curiosity, commitment, excellence, and integrity to the work I do and the people I work with. Beyond the lab, you will usually find me swimming, working out, hiking, playing badminton, exploring a new place, or stopping to photograph a view worth remembering.
Experience
Aug 2024–present · Urbana-Champaign
I work on multimodal scientific fact verification, controllable image-to-video generation, and agentic reasoning, with a broader interest in diffusion and flow models for image and video generation.
Jun–Aug 2026 · McLean
Built a distributed reinforcement-learning framework for declarative policy learning, combining Ray RLlib infrastructure with a familiar, developer-focused interface.
Jun–Aug 2025 · Atlanta
Gathered a multimodal email dataset, investigated NLP prompting methods to build a prototype evaluation pipeline, and distilled email-design heuristics into high-integrity, interpretable features.
Virginia Tech · 2021–2024
Contributed to NSF-supported research at the Laboratory for Fluid Dynamics in Nature, applying machine learning to robotic insect-wing calibration and patient-specific microfluidic insulin-pump design. I also taught programming, data structures, multimedia, and machine learning.
Selected publications
Vedant Shah, Onkar Susladkar, Tushar Prakash, Kiet A. Nguyen, Tianjiao Yu, Adheesh Juvekar, Muntasir Wahed, Ismini Lourentzou
Graph-conditioned video generation for precise, interaction-level control of multiple subjects.
Onkar Susladkar, Dong-Hwan Jang, Tushar Prakash, Adheesh Juvekar, Vedant Shah, Ayush Barik, Nabeel Bashir, Muntasir Wahed, Ritish Shrirao, Ismini Lourentzou
An inversion-free, multi-reward framework for steering diffusion and flow-matching models at inference time.
Tianjiao Yu, Vedant Shah, Muntasir Wahed, Ying Shen, Kiet A. Nguyen, Ismini Lourentzou
Part-aware 3D Gaussian representations for high-fidelity, physically consistent articulated digital twins.
Adheesh Juvekar, Onkar Susladkar, Kiet A. Nguyen, Muntasir Wahed, Nabeel Bashir, Xiaona Zhou, Tianjiao Yu, Vedant Shah, Ismini Lourentzou
A unified training-free framework for instruction-guided and subject-guided video editing.
Tianjiao Yu, Vedant Shah, Muntasir Wahed, Kiet A. Nguyen, Adheesh Juvekar, Tal August, Ismini Lourentzou
Structured elicitation and execution policies for confidence calibration in open-ended embodied environments.
Xavier Plemmons, Vedant Shah, Ismini Lourentzou
An empirical study of how data users and practitioners understand and assess data quality.
Vedant Shah
A study of SVGD and SGLD for improving the calibration and robustness of joint energy models.
Let’s connect
I’m always glad to connect with fellow researchers and builders.