Luminary has introduced Autopilot, a system that automates steps like CAD ingestion and meshing. The new capability equips agents to handle data generation, training and inference with a single prompt, even delivered via Slack. “Up to this point, setting up a Physics AI workflow required a dozen manual steps spread out over weeks or months,” said Juan Alonso, Chief Technology Officer at Luminary. The company’s Autopilot also incorporates human-in-the-loop approval checkpoints for data changes and compute resource use.
Luminary Autopilot Automates Physics AI Workflows, Reducing Time-to-Value
The system uses a new Luminary Data Lake, a training-ready store indexing platform job data, third-party simulations and reference datasets, allowing customers to train or fine-tune agents using existing and public data sources. This centralized data access is directly integrated with Luminary Pipelines, enabling users to codify workflow steps for both human and agent-initiated tasks, increasing efficiency and consistency, the company says. “Empowering engineers to automate design work simply by describing it to an agent in a single prompt is a significant milestone,” Alonso continued, emphasizing the potential for faster product development cycles.
Suds Menon, Chief Product Officer at Luminary, added that customers are demanding that manufacturing organizations deliver new innovations faster and more cost-effectively, and that Physics AI is already accelerating design timelines, reducing prototyping expenses and minimizing uncertainty in field performance. The introduction of Luminary Autopilot and Continual Large Physics Models, according to Menon, will further amplify customer development capabilities, solidifying Luminary’s position as a leader in AI-driven research and development.
Customers are demanding that manufacturing organizations deliver new innovations faster and more cost-effectively. Physics AI is helping us accelerate our design time, reduce physical testing and prototyping costs, and reduce field performance uncertainty.
Suds Menon, Chief Product Officer at Luminary
Continual Large Physics Models Enable Self-Improving, Agile Simulations
Luminary has introduced Continual Large Physics Models, a new class of physics AI that autonomously improves its predictive capabilities without requiring complete retraining from scratch. Unlike traditional models with fixed abilities after initial training, these models identify their own weaknesses, generate necessary data and learn from it, all while retaining existing knowledge. This self-improving functionality promises to accelerate engineering design cycles and reduce the need for extensive human intervention in model refinement.
The ability of these models to learn continuously addresses a key limitation of earlier physics AI systems. Previously, adapting a model to new applications or data types necessitated a full retraining process, a computationally expensive and time-consuming undertaking. Continual LPMs circumvent this by incrementally enhancing their performance through autonomous data generation and learning, effectively becoming more accurate and versatile with each use.
This approach builds on the existing capability of LPMs to predict performance characteristics early in the design cycle, now augmented by a dynamic learning process. The company is positioning these advancements as a solution to the growing demand for faster and more cost-effective innovation within manufacturing organizations, according to Luminary. The shift towards self-improving models represents a move away from static AI tools towards adaptive systems capable of evolving alongside changing design requirements and data availability.
The implications of this technology extend beyond simply reducing computational costs. By automating the model refinement process, Luminary aims to make advanced physics AI accessible to organizations with limited expertise, enabling them to benefit from its capabilities. This is particularly relevant in complex engineering fields like aerospace and automotive, where accurate simulations are important for optimizing designs and minimizing risks. The company’s approach allows for a more agile and responsive development process, potentially shortening the time it takes to bring new products to market.
Up to this point, setting up a Physics AI workflow required a dozen manual steps spread out over weeks or months and across multiple people. Every handoff introduced the chance for mistakes and took engineering resources away from critical projects.
Juan Alonso, Chief Technology Officer at Luminary
Source: Direct text input




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
