Seemandhar Jain
[email protected]
About Me
I am Seemandhar Jain, a PhD student in Computer Science at the University of California San Diego, advised by Prof. Manmohan Chandraker. My main goal is automating scientific discovery in physical AI: building research agents that find what a robot cannot yet do, design experiments that separate competing explanations, and turn the findings into new control and learning methods. Alongside this, I work on world models and policy learning for embodied agents, and I am exploring mental models for better multi-agent communication between research agents. This builds on my background in NeRFs, 3D Gaussian Splatting, and diffusion models for 3D-aware image/video synthesis.
My recent work includes Nerfify (CVPR 2026 Highlight), a multi-agent system that turns NeRF research papers into executable code; Splatify, its 3DGS counterpart with knowledge discovery; SplatLab (NeurIPS 2026), an autonomous research system that goes from literature to a new method to a validated paper; and RoboMetaHarness, which brings this research loop to robotics by automatically discovering harnesses for frozen robot policies.
Previously, I completed my MS in Computer Science at the University of Illinois Urbana–Champaign (GPA 3.92/4), advised by Prof. David Forsyth, where I worked on single-shot 3D reconstruction, convex decomposition, intrinsic image decomposition and relighting with diffusion models, and 2D-to-3D transformations. I am a Siebel Scholar, Class of 2024.
I graduated with a B.Tech in Computer Science and Engineering from the Indian Institute of Technology Indore. I founded HealthAIgnite, a healthcare AI consulting company focused on computer vision. I have also worked with industry research and product teams, including Google (Student Researcher, BigML), Aireal (Research Scientist, CV), DOCOMO Innovations (Research Intern), and Salesforce (AMTS, AI).
Quick links: LinkedIn · GitHub · Website
Research Interests
- Automated Scientific Discovery for Physical AI: research agents that reproduce baselines, form hypotheses, run experiments that separate competing explanations, and turn findings into new robot control and learning methods (SplatLab, RoboMetaHarness)
- World Models and Embodied Policy Learning: learning predictive models of the world and using them to train and plan policies for embodied agents
- Multi-Agent Research Systems: agents that read papers, implement them, and validate new methods (Nerfify, Splatify), and mental models for better communication between agents
- 3D Vision and Generative Models: NeRFs, 3D Gaussian Splatting, single-shot 3D reconstruction, relighting, and diffusion models for 3D-aware image/video synthesis
News
- [Sep 2026] SplatLab (Who Wrote This Paper?) accepted at NeurIPS 2026!
- [Jun 2026] Started as a Research Scientist Intern at Adobe Research, working on world models.
- [Jun 2026] Passed my PhD Research Exam at UC San Diego; my research now focuses on world models, policy learning, and automated research.
- [Apr 2026] Nerfify selected as a CVPR 2026 Highlight!
- [Mar 2026] Invited talk at Google on NERFIFY.
- [Jan 2026] Invited talk at Pixel Cafe.
- [Jun 2025] Student Researcher (BigML) at Google, New York; optimized SANA diffusion for Veo and built Gemma 2–based captioning.
- [Jul 2024] Started PhD in Computer Science at UC San Diego (advisor: Prof. Manmohan Chandraker).
- [May 2024] Joined Aireal as Research Scientist (Computer Vision) on single-image-to-3D textured mesh and AI interior design (Livvy).
- [Sep 2023] Selected as Siebel Scholar, Class of 2024
- [Aug 16, 2023] Started working as a Teaching Assistant for CS124: Intro to Computer Science (Fall 2023)
- [May 15, 2023] Joined the AI team in Docomo Innovations, Inc as a Research Intern!
- [Jan 17, 2023] Started working as a Teaching Assistant for CS128: Intro to Computer Science (Spring 2023)
- [Aug 22, 2022] Started working as a Teaching Assistant for CS225: Data Structures (Fall 2022)
- [Aug 15, 2022] Joined the MS CS program at University of Illinois, Urbana-Champaign!
- [Sep 3, 2021] Interviewed by Times of India for research on wildfire detection algorithm.
- [Jun 5, 2021] Joined the BI team at Salesforce as a Software Developer.
- [May 15, 2021] Graduated with B.Tech degree in Computer Science from IIT Indore.
- [May 2021] Received the Best BTP award from Computer Science Department among the graduating students.
- [Aug 2020] Secured 8th rank in IEEE Programming League hosted by IEEE Computer Society on Codingblocks.
- [Oct 2017] Secured 3rd rank in Intra IIT Tech Meet Robo Race.
- [July 2017] Achieved All India Rank 1st in IIT JEE Advanced Mathematics Section with a score of 122/122 and overall rank of 1400 among a million students.
- [May 2016] Secured AIR 43 in Technothlon hosted by IIT GUWAHATI.
- [Dec 2015] Secured State rank 2 and International rank 26 in International Mathematics Olympiad.
- [Aug 2013] Became a UCMAS degree scholar, and secured 9th rank in UCMAS global competition.
Publications
-
In Submission
Manuscript in submission, 2026. -
NeurIPS 2026
Conference on Neural Information Processing Systems (NeurIPS), 2026.PDF Project Page Accepted at NeurIPS 2026 -
3DV 2027
International Conference on 3D Vision (3DV), 2027 — under review.Code Project Page Under Review -
CVPR 2026
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026.Project Page CVPR 2026 Highlight ✦ -
ICCV-W
ICCV 2025 Workshop on Wild 3D: 3D Modeling, Reconstruction, and Generation -
NeurIPS
-
CVPR
The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR), 2024. -
CVPR
The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR), 2024.Under Review -
arXiv
arXiv preprint arXiv:2307.03847 -
ACM ToCH
ACM Transactions on Computing for Healthcare, 2022. -
IEEE TETCI
IEEE Transactions on Emerging Topics in Computational Intelligence, 2021. -
BDA
9th International Conference on Big Data Analytics, 2021. -
INDICON
IEEE 18th India Council International Conference, 2021. -
PReMI
International Conference on Pattern Recognition and Machine Intelligence, 2021.
Services
Leadership Roles
- Founded and served as CEO of HealthAIgnite, guiding strategic growth and product innovation in AI for healthcare.
- Conducted groundbreaking research and published in ACM Transactions on Computing for Healthcare.
- Managed and expanded a global team at HealthAIgnite, securing funding and partnerships, and significantly impacting the healthcare industry.
- Captained University Debate Team, achieving regional and national victories.
- Led "Fluxus" 2021 at IIT Indore amidst COVID-19, demonstrating robust leadership.
- Chaired student council at IIT Indore, managed student welfare and event coordination.
Teaching
- Graduate Teaching Assistant (UIUC): CS 225 Data Structures (Fall 2022)
- Graduate Teaching Assistant (UIUC): CS 128 Introduction to Computer Science II (Spring 2023)
- Graduate Teaching Assistant (UIUC): CS 441 Applied Machine Learning
- Teaching Assistant (UIUC): CS 124 Introduction to Computer Science (Fall 2023)
Volunteering and Leadership Experience
- Coordinated the Indian Graduate Students Association’s Holi festival at UIUC for 1000+ attendees; led community-building initiatives.
- Led resource development for HOPE NGO in India, enhancing child welfare; earned MP Young Philanthropist Award.
- Active member of AVANA club for four years, organized events for community betterment during undergrad.
- Assisted Zooniverse's solar wind turbulence projects, aiding scientific research.
Affiliations
Highlights
- SplatLab (NeurIPS 2026): autonomous scientific discovery system for 3D Gaussian Splatting research
- Nerfify (CVPR 2026 Highlight): multi-agent system that autonomously converts NeRF papers to executable code
- Accelerated large-scale video generation: 2× faster SANA diffusion sampling for Veo; Gemma 2–based captioning improved FID by 20%
- HealthAIgnite: founded and leading a healthcare CV consulting startup; partnerships across healthcare and MedTech
Honors & Awards
- Siebel Scholar, Class of 2024 (UIUC)
- Best BTP award, IIT Indore (2021)
- International Mathematics Olympiad: State Rank 2, International Rank 26 (2015)
- ACM ICPC onsite qualifier (2019); Technothlon AIR 43 (2016); UCMAS global rank 9 (2013)
Media Coverage
Gallery
A few moments with the people who shaped the journey.
How Embodied Agents Discover Robot Skills
A closed-loop robotic coding agent. A manipulator on the sim floor runs a code-as-policy program while the agent loops Execute → Inspect → Repair → Validate over its per-primitive multimodal traces. Each validated repair is distilled into a shared Skill Library — real skills like handover(), open_drawer(), can_in_trash(). Drag to orbit · click a stage to dispatch an actor.
- 1 · Execute — runs the code-as-policy program on the robot, logging per-primitive traces (perception, planning, grasping).
- 2 · Inspect — localizes the failing primitive by diagnosing the multimodal traces.
- 3 · Repair — synthesizes a targeted code-as-policy fix (e.g. a multi-angle approach).
- 4 · Validate — re-executes closed-loop; a passing repair becomes a reusable skill.
