Julia began in 2009 when a group of MIT researchers grew frustrated with programming languages used for scientific computing. Existing options could be easy to use but slow, forcing researchers to rewrite successful programs in faster languages. Their solution was Julia, a programming language intended to offer the accessibility of Python or MATLAB with performance comparable to C, tells MIT News.
The project became the Julia Lab at MIT’s Computer Science and Artificial Intelligence Laboratory. Its creators, including Viral Shah, Alan Edelman, Jeff Bezanson, and Stefan Karpinski, announced Julia publicly in 2012. Researchers quickly embraced it, and the free, open-source language now has more than one million users across universities and companies worldwide.
Julia achieves much of its performance through just-in-time compilation, which compiles code according to the type of data being processed. Its combination of speed, flexibility, and high-level abstractions has made it useful across engineering and science. Applications include modeling semiconductors, neural networks, aircraft, climate systems, brain activity, and financial markets.
Growing demand for support led the researchers to establish JuliaHub in 2015. Julia-based software has since contributed to projects ranging from pharmaceutical modeling used during Moderna’s Covid-19 vaccine development to an aircraft collision-avoidance program that ran about 50 times faster than an earlier Python implementation.
JuliaHub is now extending the language into autonomous engineering with Dyad 3.0, an AI platform for designing complex physical systems such as aircraft, rockets, satellites, and heat pumps. Engineers can direct AI agents through simulations, safety analyses, and quality-control tasks.
Unlike general-purpose AI, Dyad incorporates a physics compiler that identifies solutions violating physical laws and guides agents toward physically valid results. JuliaHub believes this combination of AI, physics, and high-performance computing could compress months of engineering design work into hours.
