Dyad: A New Language to Make Hardware Engineering as Fast as Software
a year ago
- #devops
- #hardware-engineering
- #scientific-machine-learning
- Dyad is a new modeling system designed to bridge the gap between hardware engineering and modern software development practices.
- It combines declarative physical modeling with GUI views, enabling both textual and graphical representations for AI and DevOps integration.
- Dyad integrates modern DevOps tools, including a package manager with git repositories, dependency tracking, and CI/CD pipeline compatibility.
- Built on the Julia programming language, Dyad leverages the SciML ecosystem for differentiable programming, neural networks, and GPU support.
- It supports physical modeling through acausal DAE formalisms and embedded hardware targeting for control systems.
- Dyad libraries are designed to work with generative AI, improving predictive performance and facilitating information translation from other sources.
- The system includes cloud compute integration for scalable hardware access, IoT connectivity, and digital twin development.
- Dyad Studio, a VS Code extension, is released under a Source Available license, with BSD3-licensed standard libraries for community contributions.
- Future updates will cover Dyad's tech stack, SciML methods, LLM leverage, and embedded hardware targeting.