JuliaHub raised $65 mn in Series B funding to accelerate development of Dyad, its AI platform for industrial system design, simulation, and digital engineering.
Dorilton Capital led the round, with participation from General Catalyst, AE Ventures, and former Snowflake CEO Bob Muglia. The company says several Fortune 100 firms already use Dyad and Julia across aerospace, automotive, government, utilities, and HVAC projects.
JuliaHub is a Scientific AI startup, and its mission is to empower those tackling the world’s toughest scientific and technical challenges with cutting-edge AI-first tools in a seamless, secure environment.
The company was founded in 2015 by the creators of Julia, the high-performance open-source language developed at MIT and now used by over a million developers worldwide.
JuliaHub wants engineering teams to move through design, testing, validation, and production workflows with autonomous AI agents handling large portions of the process.
The company says Dyad compresses engineering cycles from months into minutes across systems ranging from satellites and semiconductors to heat pumps and industrial infrastructure.
Physical engineering remains one of the largest sectors where AI adoption still moves slower than software development.
Coding assistants transformed software workflows fast, but industrial engineers continue working through fragmented modeling environments, disconnected simulations, and legacy infrastructure tools built long before modern AI systems arrived.
McKinsey estimates global infrastructure investment needs will reach roughly $106 tn through 2040.
According to Beinsure analysts, industrial AI companies now see a large commercial opening around engineering productivity because infrastructure expansion increasingly collides with labor shortages, system complexity, and longer design cycles.
Dyad positions itself as an AI-native engineering environment for physical systems. JuliaHub describes the platform as a software-style development workflow adapted for industrial engineering rather than conventional coding.
The company launched Dyad 1.0 in June 2025 and Dyad 2.0 in December 2025. Dyad 3.0 expands the platform with autonomous agents connected directly to scalable physics simulations, control systems, safety analysis tools, and embedded code generation workflows.
Engineers can model digital twins, adjust control systems, test deployment scenarios, and iterate hardware designs inside a single environment without switching tools repeatedly across disconnected software stacks.
Viral Shah, CEO of JuliaHub, said the company is moving beyond isolated productivity tasks toward agentic engineering at scale. According to Shah, teams can feed full specifications into Dyad and receive complete system designs as output.
The platform also uses Scientific Machine Learning, or SciML, to connect streaming operational data with evolving simulation models.
JuliaHub says its cloud-based agents continuously scan scientific research and operational inputs to improve simulations over time.
AI-automated lab testing plays a role as well. Models update as systems gather new real-world information, allowing engineers to refine assumptions, verify operational behavior, and monitor safety constraints during deployment cycles.
That safety layer matters more in physical systems than in software environments. A coding mistake breaks an application.
A flawed industrial model creates much larger problems, from equipment failure to infrastructure accidents.
JuliaHub argues general-purpose AI systems struggle in physical engineering because they cannot reliably follow the laws of physics.
The company says Dyad’s modeling language was built specifically for AI agents to reason through physical behavior, including fluid movement, thermal changes, gravity effects, and machine interactions.
According to JuliaHub, recent benchmarking in chemical process modeling showed large language models such as Codex, Claude Code, and Gemini struggled to complete early setup stages.
Dyad reportedly automated most of the process for developing model-predictive controllers inside a chemical plant optimization workflow, a task engineers often spend weeks completing manually.
David Joyce, former CEO of GE Aviation and vice chair of GE, said industrial engineering software is entering a major transition as companies search for systems combining physics modeling, control algorithms, AI, and digital twin infrastructure inside unified workflows.
Dyad also integrates with Synopsys simulation software through Ansys TwinAI. Prith Banerjee, senior vice president of innovation at Synopsys, said the platform combines scientific AI, agentic modeling, and digital twin workflows inside a single engineering environment.
One example already moved into production environments. Working with water management company Binnies and Williams Grand Prix Technologies, JuliaHub developed a SciML-powered digital twin capable of predicting pump faults in water distribution systems using only four sensor inputs. According to the company, the system achieved more than 90% prediction accuracy.
Tom Ray, director of digital products and services at Binnies, said the platform shifts water infrastructure management from reactive operations toward predictive system-level decision-making.
The broader market now watches whether AI-generated engineering systems move beyond pilot programs into large-scale industrial deployment.
Software automation spread fast because digital mistakes were easier to fix. Physical engineering doesn’t offer that luxury. JuliaHub is betting physics-aware AI models finally narrow that gap enough for industrial firms to trust autonomous engineering workflows at scale.
