September 2026 Summaries

Rescale introduces two Assistant-powered approaches to Design of Experiments that expand on its traditional DOE job type, which supports simple parameter sweeps but has limited algorithms and lacks Custom Fields, Automations, and AI Physics integration. Rescale Assistant can generate DOE matrices using methods including full and fractional factorial designs, Latin hypercube sampling, Box-Behnken, central composite design, and Monte Carlo sampling, then execute them through Workflows. The Workflow Runs approach creates a separate independent workflow for each design point, supporting flexible pre- and post-processing, simultaneous launches, fault isolation, and reruns of individual failed cases. Alternatively, a single Workflow Template places all design points on one canvas, enabling engineers to inspect configurations and dependencies before submission, centrally monitor progress, control concurrency, and retain one complete record of the study. Both options work with any Rescale-supported solver and make results available through Custom Fields for analysis or AI Physics model training.

Engineering organizations are approaching a new phase of digital transformation in which agentic AI systems could extend beyond basic content and coding assistance to coordinate end-to-end engineering workflows across CAD, CAE, PLM, simulation, testing, manufacturing, and business operations. Rescale argues that early adopters can create a compounding advantage by capturing institutional knowledge, reducing iteration costs, and accelerating learning across programs, provided agents are grounded in company-specific engineering practices, solver preferences, quality standards, and review processes. Proposed uses include automating simulation studies and surrogate-model development, coordinating multi-physics analyses, identifying the downstream effects of design changes, assessing compliance and design quality, correlating simulation with physical tests, generating RFQ responses, detecting manufacturing risks, and supporting life-sciences and chemical scale-up. The approach relies on an engineering knowledge layer, workflow orchestration tools, integrations with engineering software and computing resources, and built-in governance such as approval gates, audit trails, domain-specific evaluation, and human oversight. Rather than immediately pursuing full autonomy, organizations are encouraged to begin with a high-value, measurable workflow and gradually expand delegation as confidence in agent performance grows.

Rescale is introducing AI Physics Auto Tune, a forthcoming feature intended to simplify surrogate model development and deployment for engineering applications. Users can set a search duration while the tool evaluates hyperparameter configurations in parallel, eliminates weaker candidates early, fully trains leading options, and selects the best-performing model from four production-ready architectures. The feature is designed to help users develop faster and more accurate AI physics models without requiring extensive AI or machine learning expertise.

Rescale has introduced HiLiftAeroML, an open-source aerodynamic dataset available within its AI Physics platform to help engineers develop AI-based physics and surrogate models. Derived from surface simulation data covering four geometries and multiple angles of attack, the dataset provides a starting point for model training without requiring users to first create and assemble their own data. The announcement emphasizes its potential to accelerate adoption of AI physics workflows and encourages users to explore Rescale AI Physics for building custom surrogate models with open datasets.

A September 2026 Rescale Innovation Corner post by Viktor Rozsa describes an agentic multi-disciplinary engineering workflow using Ansys CFX and Mechanical. The approach combines deterministic simulation with agentic engineering to review completed CFX design studies, identify influential parameters and missing data, initiate follow-up analyses, and transfer CFX pressure loads into Ansys Mechanical. By connecting hydraulic and structural results, the workflow is intended to support more integrated downstream engineering decisions on the Rescale platform.

Viktor Rosza describes an agentic multi-disciplinary engineering workflow on Rescale that integrates Ansys CFX computational fluid dynamics with Ansys Mechanical structural analysis. The approach uses agentic engineering alongside deterministic simulation to review completed CFX design studies, identify influential parameters and missing data, initiate follow-up analyses, and transfer CFX pressure loads into Mechanical. By connecting hydraulic and structural results, the workflow aims to help engineers evaluate downstream design decisions across disciplines more efficiently.

Rescale has expanded its job submission environment to support multi-node MPI and Kubernetes-based workloads on CoreWeave Cloud, aiming to provide enterprise HPC capabilities at potentially lower costs for large-scale simulations. The company reports that the integration has been validated across multi-node configurations through scale testing and initial enterprise production use, and encourages organizations interested in Kubernetes orchestration and specialty-cloud cost optimization to contact Rescale.

Rescale has expanded its job submission environment to support multi-node MPI and Kubernetes-based workloads on CoreWeave Cloud, aiming to provide enterprise HPC capabilities at lower costs for large-scale simulations. The integration has been validated through end-to-end multi-node testing, scale tests, and initial enterprise production use, and Rescale encourages organizations interested in Kubernetes orchestration or specialty-cloud cost optimization to contact the company.