Blog | Luke Hollis

2025

Genie-Bottle: Interpretable Generative Interactive Environments

PyTorch implementation of generative interactive environments with discrete latent action codebooks for interpretable control.

L Luke Hollis
November 23, 2025
1 min read

generative-models pytorch world-models interpretability
LLM-Driven Multi-Agent Urban Simulations

Generalist agent training framework for game engines using hybrid RL/ML approach for urban planning and logistics simulation.

L Luke Hollis
November 23, 2025
1 min read

reinforcement-learning urban-simulation multi-agent llm
Caduceus: Open-Source AI Medical Scribe

Real-time medical documentation system with voice transcription and clinical template generation.

L Luke Hollis
November 23, 2025
1 min read

healthcare ai nlp medical-documentation
SF 20K Personas: Large-Scale Synthetic Population Dataset

20,000 procedurally generated personas with demographics, behaviors, and social networks for San Francisco urban simulation.

L Luke Hollis
November 23, 2025
1 min read

synthetic-data urban-simulation demographics social-networks
SPHR: Spherical Harmonics Reconstruction

Spherical harmonics-based 3D reconstruction and rendering system for volumetric data representation.

L Luke Hollis
November 23, 2025
1 min read

3d-reconstruction rendering computer-graphics spherical-harmonics
Text-to-3D Simulation on Map: Language-Driven Geospatial Modeling

Natural language interface for generating and placing 3D simulations on geographic maps using LLM-driven scene generation.

L Luke Hollis
November 23, 2025
1 min read

3d-generation geospatial nlp simulation
Three-MLAgents: Unity ML-Agents for Three.js

Port of Unity ML-Agents to Three.js enabling browser-based reinforcement learning training and deployment.

L Luke Hollis
November 23, 2025
1 min read

reinforcement-learning three-js webgl ml-agents