Julia programming language grows from MIT project to 1M+ users

Julia began in 2009, when a group of researchers frustrated with the rigid, slow programming languages used for scientific computing started venting over email. Those older languages forced scientists to rewrite a whole working program in a different language just to make it run fast enough. The emails turned into a research project at MIT aimed at building an easy to use, high performance language for scientific research, data analysis and modeling complex systems such as jet engines, drugs, financial markets and robots. Around the same time, Julia's co-creators started the Julia Lab, still active today inside MIT's Computer Science and Artificial Intelligence Laboratory. 'We wanted to create something as easy to use as Python or MATLAB but as fast as the C programming language,' says JuliaHub co-founder and CEO Viral Shah. 'We built Julia for ourselves.' Fellow co-creator Alan Edelman, an MIT mathematics professor, recalls that the team did not expect much interest at first: they figured it would take 10 years before anyone cared, but decided patience was a virtue and pressed ahead anyway.

MIT researchers announced Julia publicly with a blog post in 2012, and quickly found that many other researchers shared their frustration. 'When we first started, we were targeting interactive research workflows, but increasingly people are using it for everything,' says co-creator Jeff Bezanson. 'Now we're moving the whole stack of the language onto smaller, embedded devices as we evolve with our users.' Edelman has taught a Julia class at MIT with students from nearly every department, and says he often finds students are already using the language for robotics, astronomy, physics simulations and finance before they even enroll. He credits Julia's abstractions: 'A lot of times a coding language forces you to solve the one problem you're thinking about. Julia's language makes it so you're solving not only the problem you're thinking about, but other people's problems around the world too. It encourages you to solve problems more generally.'

As Julia's popularity grew, researchers worldwide began asking the team for support, and by 2015 demand was strong enough that the team started a company, JuliaHub, to help users full time, with support from the MIT Deshpande Center for Technological Innovation and others at MIT. JuliaHub's founders include Julia's co-creators: CEO Viral Shah, Alan Edelman, Jeff Bezanson (MIT SM '12, PhD '15), and former MIT research scientist Stefan Karpinski.

Today, Julia is free and open source and counts more than 1 million users across thousands of companies and universities. Its edge, according to the piece, is that it compiles code differently depending on the type of data involved, a technique called just in time compilation that the article says makes Julia faster and more flexible than other numerical programming languages. That design reflects the founders' original goal, in Shah's words: 'Scientists and engineers are not programmers. Building scientific applications with multidisciplinary teams of scientists, engineers, and programmers is challenging. We asked: What if you could equip the scientists and engineers with a programming language that allowed them to express their ideas at a high level and also get great software performance?'

The piece calls it only a slight exaggeration to say Julia has been used to model everything under the sun: tiny atoms, semiconductors, neural networks, race cars and airplanes, and, beyond Earth, black holes that astronomers have imaged using it. It has also been used to simulate computer circuits, detect health disparities, model global climates and oceans, and analyze brain activity. Concretely: someone built a pharmaceutical modeling platform in Julia (the article does not say who) that was later used to help speed up development of Moderna's Covid-19 vaccine. Researchers built a Julia program for avoiding aircraft collisions that they found ran about 50 times faster than an earlier version built in Python. And Meta engineers used Julia to build a better audio codec for WhatsApp, which has 4 billion users.

JuliaHub's newest push is Dyad, an AI platform meant to help engineering teams accelerate the development of complex physical systems such as rockets, heat pumps and satellites, with engineers directing autonomous AI agents through physics simulations, safety analyses and quality controls. JuliaHub launched Dyad 1.0 in June 2025 as a research agent to accelerate programming, followed by Dyad 2.0 that December and Dyad 3.0 in April; the founders say Dyad 3.0 marks a new level of ability and autonomy for designing complex systems. Shah describes Dyad as fundamentally 'a physics compiler,' meaning it enforces physical laws: general AI systems, he says, often solve physical problems in ways that violate those laws, and Dyad is built to catch such violations and steer the process toward a physically correct solution. He says the company expects this to cut product engineering design times by orders of magnitude, turning months of work into hours, though that figure is presented as the company's own expectation rather than a measured result. 'With Dyad 3.0, you can upload data and design documents and the system will design an entire aircraft for you,' Shah says, citing Boeing as an example of the customers JuliaHub is working with on this kind of agentic hardware design. He frames the ambition simply: tell the system to build something (his own example is 'build me a plane'), upload the design documents, and have Dyad account for all the physics, compile the code, verify everything and produce the full design agentically.

Edelman says he is already seeing the effect in his own class: one student recently used Dyad to model how robots move in space, and another used it to build a rocket engine, telling Edelman afterward that he could hardly believe how easy it had been to get a working rocket engine out of it.

Key facts

Julia began in 2009 when a group of researchers frustrated with slow, rigid programming languages for scientific computing started venting over email; the project became MIT's Julia Lab and was announced publicly via a blog post in 2012.
By 2015, demand for support had grown enough that Julia's creators started the company JuliaHub, with backing from the MIT Deshpande Center for Technological Innovation; its founders are CEO Viral Shah, MIT mathematics professor Alan Edelman, Jeff Bezanson (MIT SM '12, PhD '15), and former MIT research scientist Stefan Karpinski.
Julia is free and open source and today counts more than 1 million users across thousands of companies and universities, with cited examples including a Julia-based pharmaceutical modeling platform later used to help speed up Moderna's Covid-19 vaccine development and an aircraft-collision-avoidance program its developers say ran about 50 times faster than an earlier version built in Python.
Meta engineers used Julia to build a better audio codec for WhatsApp, which has 4 billion users.
In April, JuliaHub launched Dyad 3.0, the newest version of an AI platform (after Dyad 1.0 in June 2025 and Dyad 2.0 that December) that Shah calls a 'physics compiler' meant to catch AI-proposed designs that break physical laws; citing Boeing as an example customer, Shah says the goal is for engineers to upload design documents and have autonomous agents handle the physics, verification and hardware design.

Why it matters

Most academic programming languages stay in the lab. Julia did not. It grew from a 2009 round of frustrated emails among MIT researchers into a free, open-source language with more than 1 million users, and into a company, JuliaHub, built around it. That is the rarer trajectory: an academic tool actually crossing into mainstream, industrial use. The same team is now trying to repeat the trick one level up. Instead of a language that lets scientists express high-level ideas while still getting fast code, JuliaHub's Dyad platform aims to let engineers describe a physical system in high-level terms and have AI agents handle the low-level physics, verification and design work. Dyad 3.0, launched in April, is pitched by JuliaHub as a step change in how autonomous that process can be, and as what Shah calls a 'physics compiler': a system meant to keep AI-generated designs inside the bounds of real physical law, unlike what he describes as general AI systems that can propose physically invalid solutions.

Who it affects

Directly: scientists, engineers and mathematicians who need high performance numerical computing without hand optimizing their own code, a group the piece says spans thousands of companies and universities. Concrete examples cited include Moderna, whose Covid-19 vaccine work was sped up by a Julia-based pharmaceutical modeling platform; Meta, whose engineers used Julia to build a better audio codec for WhatsApp's 4 billion users; and the team behind a Julia-based aircraft-collision-avoidance program. For Dyad specifically: engineering teams building complex physical systems such as rockets, heat pumps, satellites and aircraft, with Boeing named as an example customer JuliaHub is working with on agentic hardware design. MIT students are affected too. Edelman says students in his Julia class increasingly arrive already using the language, and some are now using Dyad for class projects, in one case to model how robots move in space and in another to build a rocket engine.

How to use it

Julia itself is free and open source. For Dyad, the workflow Shah describes is to upload data and design documents and let the system design the physical system for you, in his words 'an entire aircraft.' He frames the broader ambition as being able to tell the system to build something (his own example is 'build me a plane'), upload the relevant design documents, and have Dyad account for the physics, compile the code, verify everything and produce the full design agentically. Today, engineers use Dyad to direct autonomous AI agents through physics simulations, safety analyses and quality controls, and JuliaHub cites Boeing as one customer it is working with on this kind of agentic hardware design.

How solid is it

This is MIT's own news office writing about work done at MIT, built largely on quotes from Julia and JuliaHub's own founders (Shah, Edelman, Bezanson), so it reads as an institutional retrospective rather than independent outside reporting. The historical timeline (the 2009 origins, the 2012 announcement, the 2015 founding of JuliaHub, and the Dyad 1.0, 2.0 and 3.0 launches) and the named use cases (Moderna, Meta, the aircraft-collision-avoidance program) are presented as established fact. The forward-looking claims about Dyad read softer. The founders say they believe Dyad 3.0 marks a new level of ability and autonomy, and Shah's expectation of an 'orders of magnitude' cut in design time, turning months of work into hours, is explicitly framed as the company's own expectation, with no specific numeric factor given. His 'upload documents and get an entire aircraft' framing and his 'build me a plane' example are likewise his own hypothetical illustrations of the goal, not a description of a completed project; Boeing is named only as an example customer JuliaHub is working with on agentic hardware design, with no further detail on that relationship given in the piece.

Risks and caveats

The piece does not name who built the pharmaceutical modeling platform used on Moderna's vaccine work, or give dates for when the Moderna, Meta or aircraft-avoidance projects took place. It also gives no revenue, funding, headcount or valuation figure for JuliaHub as a company, and no detail on the scope or timeline of the Boeing relationship beyond Shah naming it as a customer example. The clearest caveat concerns Dyad itself: its most striking claims, that it can turn uploaded documents into 'an entire aircraft' design and cut engineering time by orders of magnitude, are the founders' stated goals and expectations, not measured or independently confirmed results, and readers should weigh them accordingly.

“With Dyad 3.0, you can upload data and design documents and the system will design an entire aircraft for you.”

— Viral Shah, JuliaHub co-founder and CEO