Luke Hollis
Researcher, simulation and autonomous systems. Focus on world model development and causal discovery. Repeat founder with 10+ years leading teams that turn simulation research into software for public sector and industry.
[This website is under development, may be incomplete.]
EXPERIENCE
2025-
Researcher, Massachusetts Institute of Technology, Media Lab, City Science SF
World Models, Simulation, Autonomous Systems, Causal Inference
2025-
Research Assistant, Harvard University
LiDAR, Remote Sensing
2023–2025
Applied AI Engineer and Advisor, Startups
Machine Learning, Product Development, Technical Strategy
2020–2022
Lead 3d Research Engineer, Harvard University
3D Reconstruction,
Agent-based Modeling, Computer Vision
2019–2020
Visiting Researcher, Massachusetts Institute of Technology
Media Lab, Computational Design, Human-Computer Interaction
2014–2019
Founder, Researcher, Archimedes Digital
Product Development, Educational Technology, Archives as a Service
EDUCATION
2025-2027
Harvard University, Masters of Data Science
Harvard Kennedy School, Intelligence Project Fellow
2025
Stanford University, Graduate Coursework
Data Science and Artificial Intelligence, Applied Math
Research
MIT Media Lab, Harvard Computational Robotics, Harvard Visual Computing
ADVISORY POSITIONS
2024
Apple, iPhone Camera Team, 3d Capture
2024
IBM Almaden Research Center, Spatial Intelligence
2018-2019
Google, Artificial Intelligence (NLP) Admin, Summer of Code
OPEN SOURCE
Three.js + Torch implementation of Unity's ML-Agents framework
Agent training and visualization in browser. Python, Three.js
26
1
Multi-agent simulations in game engines
Realistic human behaviors and synthetic training data. C#
28
3
Open, hackable 3d digital twin / virtual tour builder
Supports 360 images, 3d gaussian splatting, and vfx. JavaScript
28
8
Pytorch implementation of "Genie: Generative Interactive Environments"
Forked from myscience/open-genie. Bruce et al. (2024). Python
3
2
20,000 synthetic human profiles of SF citizens
Modeled on real US Census data for use in simulations.
3
Healthcare AI scribe for real-time medical documentation
Clinical templates and real-time transcription. TypeScript
7
SELECTED PRESS
COMPUTER LANGUAGES
Languages
Python, Javascript, C#, C++, R
Technologies
Torch/Tensorflow, Three.js, Unreal, Unity, Next.js, React, foundation world models and robotics / simulation environments, Isaac Gym, Genesis/Cosmos/V-JEPA 2, MuJoCo, ArduPilot, XPlane, SITL, iNAV, etc.
Methodologies
Reinforcement Learning, Self-Supervised Learning, Causal Inference
AWARDS AND HONORS
2025
Sequoia Capital, Start-up Trek, selected founders from Harvard and MIT
2025
Artificial General Intelligence House, World Models Hackathon, 1st Place Winner
2025
Replicate, Digital Twins Hackathon, 1st Place Winner
2025
e14, Artificial Intelligence at MIT Summit, Mountain View
2024
Matterport Digital Twin Award, Top 10 Most Viewed Space of Year
2024
National USS Hornet Hackathon, 1st Place Winner
2024
Artificial General Intelligence House, Simulations Hackathon, 1st Place Winner
2023
Matterport Digital Twin Award, Honorable Mention
2020
Graduate Research Fellowship, Massachusetts Institute of Technology
SELECTED HIGHLIGHTS
- Built an open source version of Google DeepMind Genie generative world models that I presented at Google and implemented for RL training and simulation
- Sold software to every top 100 university in the world at my first startup
- Created an open-source ML Library for Three.js based on one from Unity
- Built a text-to-3d simulation platform used by major construction companies and intelligence analysts for running experiments and risk estimation
- Started and co-organize the Harvard/MIT Reinforcement Learning student group
Selected Publications
Scaling Language-Grounded Splatting to Large Urban Spaces
2026Novel approach to scale language-grounded 3D gaussian splatting techniques for large-scale urban environments.
Yasushi Sakai, Luke Hollis
MIT Media Lab & Harvard University
In Progress
3D Gaussian Splatting Computer Vision Urban Computing
Predictive Digital Twins: A World Model Approach to Deformable Mesh and Soft-Body Simulation Physics
2026Applying world model architectures to predictive physics simulation for deformable objects.
Heng Yang, Luke Hollis
Harvard Computational Robotics Group
In Progress
World Models Physics Simulation Digital Twins
Deformable LangSplat: Integrating Physics-based Segmentation for Large-Scale Radiance Fields
2025Integration of physics-based deformation with language-guided 3D gaussian splatting.
Wanhua Li, Hanspeter Pfister, Zhutian Chen, Luke Hollis
Harvard Visual Computing Group & U. Minnesota
Exploratory
Neural Radiance Fields Physics Simulation Segmentation
Urban Planning and Design with 3D Street Captures and 360 Video
2025Methodology for integrating immersive 3D captures into urban planning workflows.
Peter Hirshberg, Kate Connally, Lee Crawford, Luke Hollis
MIT Media Lab
Forthcoming
Urban Planning 360 Video 3D Capture
3D Digitization and Capture with 3D Gaussian Splatting with Color Standardization
2025Color standardization techniques for consistent 3D gaussian splatting reconstructions.
Heather Hurst, Boris Beltrán, Franco Rossi, Luke Hollis
MIT & Skidmore College
Published
3D Gaussian Splatting Color Science Digitization
Latest posts
PRESENTATIONS AND INVITED TALKS
2025
Autonomy and Robotics, Ethics and Engineering Guest Lecture, Massachusetts Institute of Technology.
Techniques and Development in Remote Sensing, Lecture, Harvard University.
3d Capture at Stanford Campus, Part 2 of 2. Apple. Stanford University.
3d Capture at Stanford Campus, Part 1 or 2. Apple. Stanford University.
Language-Conditioned Simulation with Reinforcement Learning: Synthesizing Decision-Grade Digital Twins for Complex Physical Systems. Università Cattolica del Sacro Cuore
2024
Generative Interactive Environments: World Models and Learning from GDM Genie. Google
3d Capture and Digital Generative Environments for High Fidelity Urban Simulations of Historical Spaces. Harvard University
Techniques and Methodologies in 3d capture and 3d Gaussian Splatting. Old Dominion University.
2023
Synthesizing Multiple Data Imports for 3d Captures of Large Architectural Features. Università degli Studi di Pavia
2022
Digitization and 3d Simulations for Large Exterior Architectural Spaces. Columbia University
2019
Neural Rendering: 3d Capture and Narrative Building in Extended Realities. American Research Center in Egypt
Extending Reality with Reality Capture Workshop. American University of Cairo
3d Graphics in Extended Reality. Massachusetts Institute for Technology
Building our Digital Imagination: XR in 3d Capture. Women in Games Boston
Designing across Virtual, Augmented, and Mixed Realities for 3d Capture. Harvard University
Creating Immersive Multiplayer Environments for Data Visualization. Computer Applications and Quantitative Methods
Immersive Data Visualization and Interpretation. Computer Applications and Quantitative Methods
2018
Challenges in the Creation of Digital Textual Annotation. Leonard Muellner and Luke Hollis, Harvard University
XR 3d Capture Architectural Data. Harvard University
Digital Futures Consortium: Visualizing 3d Captures in XR. Harvard University
Boston University Mellon Seminar.
Digital Preservation spanning 2D and 3D Interfaces: Cross-platform visualization of data from recording and reconstruction. American Institute Annual Conference
2017
The Harvard Yard 3d Capture Architectural Features Project. Brandon Bentley, Aden Brown, Jeffrey P. Emanuel, and Luke Hollis. Harvard University
Harvard Yard 3d Graphics Lecture, Harvard University
Cityscale handheld AR and cross-platform VR for visualizing georeferenced datasets as a part of a microservices software architecture. Computer Applications and Quantitative Methods