ConstructGPT
How can we model construction sites using diffusion, world models, and multimodal RAG?
Compare site geometry with BIM to support monitoring and compliance checks.
Read the paper ↗Multimodal LLMs · 3D vision · Agentic systems
I’m an M.S. student in Computer Science at the University of Wyoming, advised by Dr. Shivanand V. Sheshappanavar and Dr. Ahmed Abdelaty. My research connects vision, language, and 3D geometry to help AI reason about real engineering problems.
I’m fascinated by this question: How can we build AI systems that bring together multiple modalities to understand and interact with the real world?
Current research
The questions behind the workPlanSightRAG
How can AI answer questions about engineering drawings?
Retrieve visual evidence directly from plans to ground answers and compliance decisions.
Read the paper ↗WYDOT assistant
How can engineers find answers across scattered documents?
Connect engineering documents in a knowledge graph for multi-agent retrieval and reasoning.
Details in my CV ↗Selected research
All publications-
Generative modeling · EEGUnder review
MontageFlow: Physically Valid High-Fidelity EEG Generation
arXiv preprint · 2026
Experience
Research & practiceAug 2025 — Present
Graduate Research Assistant
Geometric Intelligence Research Lab · University of Wyoming
Developing multimodal RAG, 3D vision-language models, world models, and agentic AI for engineering applications.
Summer 2026
Research Collaborator
Dartmouth College
Contributed to MontageFlow, including EEG preprocessing, experiments, conditioning studies, physiological analysis, and manuscript writing.
Summer 2026
Summer Research Intern
University of Wyoming
Built a GPU-accelerated vision pipeline for traffic and parking analytics, including object tracking, occupancy detection, and speed estimation.
Apr — Aug 2024
Junior Machine Learning Engineer
PalmMind Technology · Nepal
Built production RAG chatbots for EV and insurance clients with LangChain, LangGraph, OCR, and Redis caching.
Education
2025 — 2027
(expected)
M.S. in Computer Science
University of Wyoming · GPA 4.00/4.00
2019 — 2024
B.E. in Electronics, Communication & Information Engineering
Institute of Engineering · Tribhuvan University, Nepal
Recent highlights
Graduate Scholar Award · School of Computing, University of Wyoming · $5,000 · 2026–2027
UW AI Summit, September 2026 · Lightning talk: “Teaching Models to Work in 3D: Perceiving, Predicting, and Generating 3D Infrastructure Geometry.”
Let’s connect
Ideas are better in collaboration.
I’m open to collaborations in multimodal AI and domain applications.
nsubedi1@uwyo.edu