TechPrufer
Get started
NVIDIA
NVIDIA logoNVIDIA · Feature trail

AI-Assisted Scientific Simulation Workflow Generation

3 posts · May 2026 – Oct 2026

What this trail covers

This feature thread tracks the evolution of AI-assisted tools and workflows for scientific simulation, focusing on enabling researchers to generate complex simulation code and pipelines using natural language prompts. Initial efforts focused on the NVIDIA ALCHEMI Toolkit, providing composable, PyTorch-native, GPU-accelerated building blocks for Machine Learning Interatomic Potentials (MLIP). Subsequent developments have integrated AI coding agents with ALCHEMI Toolkit agent skills and reference materials to automate the conversion of CAD assets into simulation-ready OpenUSD assets for robotics. This includes using AI models to guide the import of STEP files, match visual appearance, configure physics properties, validate against SimReady requirements, and run pick-and-place tasks in Isaac Sim. Frontier AI models like GPT-6 Astra can assist by writing Python code that calls Omniverse tools at each stage, reducing manual preparation effort. Physics configuration uses manufacturer datasheets and URDFs to define joint axes, limits, and mass properties, with contact friction coefficients. SimReady validation checks units, geometry, materials, rigid bodies, joints, drives, articulation, and runtime behavior.

All NVIDIA trails
  1. 5 Steps to Create SimReady Assets for Robotics with Frontier AI Models | NVIDIA Technical BlogThis post details a five-step workflow for converting CAD assets into simulation-ready OpenUSD assets for robotics using NVIDIA Omniverse libraries, SimReady Foundation specifications, and AI agent skills. It demonstrates the process with an ABB YuMi robot, highlighting the use of GPT-6 Astra to generate Python code for asset import, appearance matching, physics configuration (including joint axes, limits, mass properties, and friction coefficients of 0.8 static/0.6 dynamic), SimReady validation, and pick-and-place task execution in Isaac Sim. The workflow leverages the CAD-to-SimReady skill for agentic guidance and emphasizes the importance of SimReady profiles for validating asset structure, materials, and physics properties.
  2. How to Use AI Agents to Prepare 3D Scenes for Simulation | NVIDIA Technical BlogThis post details an agentic workflow for preparing 3D scenes in Blender for robotics simulation. It introduces the use of Codex (powered by GPT-6 Astra) as an orchestrator and specialized subagents built with NVIDIA NemoClaw and the Hermes agent harness. These subagents leverage NVIDIA Omniverse Libraries (ovphysx for physics, ovrtx for rendering, and SimReady for validation) to inspect scenes via Blender's Model Context Protocol (MCP), author simulation-relevant metadata in OpenUSD, configure physics properties, and render preflight views. The workflow aims to automate the tedious preparation of 3D assets for simulation environments like NVIDIA Isaac Sim or Isaac Lab.
  3. How AI Coding Agents Can Unlock Materials Simulation with NVIDIA ALCHEMI Toolkit | NVIDIA Technical BlogThis post details the integration of AI coding agents with the NVIDIA ALCHEMI Toolkit to generate end-to-end simulation workflows for materials science from natural language prompts. It introduces agent skills and reference files that bridge the gap between natural language and the ALCHEMI Toolkit API. The post presents a systematic benchmark of 45 pipelines across three workflows (Silicon EOS, O/Cu(111) adsorption, Li self-diffusion) and five prompt levels, demonstrating that prompt detail influences code structure and reusability but not physical correctness. It highlights the importance of explicit naming of materials, phases, and reference conventions, and the need for explicit requests for self-assessment and premise-checking from the AI agent. The execution of these workflows on NVIDIA H200 GPUs is also discussed.
  4. Synthesize Realistic 3D Medical Images at Scale to Ship Pre‑Trained Models | NVIDIA Technical BlogThis post introduces NV-Generate-MR-Brain, a new model for synthetic generation of human brain anatomy and structure segmentation, built on the MAISI architecture. It extends the MAISI framework towards scalable, open workflows for synthetic 3D medical imaging generation. The model is trained on the MR-RATE dataset, the world's largest open-source multimodal MRI dataset, comprising 100,000 brain MRI studies. The NV-Generate-CTMR framework, which includes NV-Generate-MR-Brain, is presented as an open-source, end-to-end solution for synthetic CT and MRI generation, supporting flexible voxel sizes, variable volume dimensions, and whole-body coverage. It emphasizes reuse over retraining for efficient and sustainable AI development.