Artificial intelligence has rapidly transformed software engineering, with generative AI and large language models (LLMs) now capable of producing vast quantities of code and documentation, while machine-learning algorithms monitor performance and detect security vulnerabilities. However, the question of whether these AI tools can be equally transformative when the task involves conceiving, designing, and fabricating complex physical systems, such as a jet engine, has remained a significant area of exploration. This past semester, the Jet-engine AI Research and Validation Intensive Sprint (JARVIS Challenge) at MIT sought to answer precisely this, by asking undergraduates to investigate whether AI could compress the traditionally lengthy design-build-test cycle, ultimately enabling them to build faster and with greater efficacy.
The groundbreaking initiative, orchestrated by MIT’s Gas Turbine Laboratory and a consortium of faculty and industry partners, concluded with a resounding affirmation of AI’s potential in safety-critical hardware engineering. Yet, it also underscored an enduring truth: engineering judgment remains the decisive differentiator. Professor Zolti Spakovszky, director of the MIT Gas Turbine Laboratory, articulated this nuanced outcome: "The JARVIS challenge showed that AI can substantially accelerate safety-critical hardware engineering, but engineering judgment remains the decisive differentiator. An AI-native engineer is not defined by using AI, but by leading it — knowing when to trust it, when to challenge it, and how to translate AI outputs into working hardware. Manufacturing — not engineering design or analysis — remained the fundamental rate-limiting step." This insight highlights a critical shift in the engineering paradigm, where human expertise evolves from merely executing tasks to skillfully directing sophisticated AI tools.
The Grand Challenge: AI in Hardware Engineering
The aerospace industry, a sector synonymous with precision and complexity, typically operates on protracted development cycles. Designing a new jet engine, for instance, can span years, even decades, involving thousands of engineers, billions of dollars, and rigorous testing protocols to ensure safety and performance. This extensive timeline is a direct consequence of the intricate interplay of aerodynamics, thermodynamics, materials science, and structural mechanics, where even minor design flaws can lead to catastrophic failures. In this context, the notion of compressing the design-build-test cycle from years to mere weeks, as explored by the JARVIS Challenge, represents a potentially revolutionary shift.
Traditional jet engine design relies heavily on iterative simulation, expert knowledge, and physical prototyping. Each stage is time-consuming and resource-intensive. The advent of generative AI and advanced computational tools offers a tantalizing possibility: could AI automate aspects of design, accelerate analysis, and identify optimal solutions at an unprecedented pace? The JARVIS Challenge was conceived as a crucible to test this hypothesis, specifically focusing on the most demanding of physical systems.
The Genesis of JARVIS: MIT’s Vision for AI-Native Engineering
The JARVIS Challenge was not merely an academic exercise; it was a strategic endeavor by MIT to pioneer the concept of "AI-native engineering." Recognizing the burgeoning capabilities of AI, particularly in generative models and LLMs, faculty leadership sought to understand their practical application in a domain historically resistant to rapid automation: complex physical hardware. The challenge aimed to equip the next generation of engineers with the skills to leverage AI as a powerful co-pilot, fostering a culture where human ingenuity and machine intelligence collaborate seamlessly.
Faculty leadership included professors Zachary Cordero, Zolti Spakovszky, Masha Folk, and Andreea Bobu from the Department of Aeronautics and Astronautics, alongside engineers from MIT Lincoln Laboratory and a dedicated team of teaching assistants. Their collective goal was not only to observe how students would integrate AI into their workflows but also to ensure the safety and educational rigor of the intense four-week sprint. Weekly progress reviews served as critical junctures where faculty would evaluate student progress and scrutinize their AI usage, offering guidance without dictating solutions. Professor Spakovszky’s subtle prompting, such as asking, "Do you know what a rabbet fit is? Take in the comment," exemplified a pedagogical approach designed to foster critical thinking and self-discovery rather than direct instruction.
Setting the Stage: Tools, Teams, and an Ambitious Task
The JARVIS Challenge offered undergraduates an unparalleled opportunity: four weeks to design, fabricate, assemble, and test a small gas turbine aero engine, with AI serving as their primary engineering partner. The ambitious objective was to construct a "JARVIS-class" single-spool jet engine capable of producing 50–100 pounds of thrust, operating on Jet-A fuel, and completing five consecutive 60-second runs. Teams were granted complete autonomy over design choices, materials selection, and fabrication methods, pushing the boundaries of conventional engineering constraints.
Thirty-one students, representing nearly every department within the School of Engineering, formed seven distinct teams. Their experience levels varied significantly, from all-first-year groups to those heavily populated by seniors. Remarkably, many participants initially possessed limited knowledge of turbomachinery, compressible flows, or even fundamental thermodynamics. For some of the younger students, this was their first encounter with the internal workings of a gas turbine. This diverse cohort provided a unique testing ground for observing how different levels of foundational knowledge interacted with AI assistance.
To facilitate their ambitious undertaking, students had access to a comprehensive suite of resources. This included MIT’s advanced machine shops and a network of manufacturing vendors, commercial engineering software such as Concepts NREC (a leading turbomachinery design suite), SolidWorks (for detailed CAD modeling), and ABAQUS (for finite element analysis). Additionally, various test rigs were available for characterizing and assembling individual components, ensuring that students could validate their designs throughout the process.
A cornerstone of the challenge’s AI integration was MIT Parley, a newly launched platform that aggregates frontier large language models through a single, intuitive interface. JARVIS leads were granted early access to Parley for all participants, enabling them to monitor AI usage patterns, including prompt frequency, cost per prompt, and the specific LLMs employed. This transparency offered invaluable insights into how students interacted with the AI tools. With substantial financial backing from MIT Lincoln Laboratory, the Department of Mechanical Engineering, and corporate sponsors like Safran, Voyager Technologies, Beehive Industries, and Boom Technology, students enjoyed essentially unlimited access to AI resources, removing any cost barriers to experimentation.
The involvement of corporate sponsors was driven by both recruiting interest and a genuine curiosity about AI’s potential to redefine engineering workflows. Ryan (Hal) Hefron of Voyager Technologies underscored the industry’s perspective, stating, "We see this as the future of engineering. You’re honing skills that are not just nice to have — they’re going to be the future baseline in the engineering workforce." Vincent Garnier, managing director of Safran Tech, echoed this sentiment, observing the students’ journey with excitement. "JARVIS was a genuine experiment, a learning endeavor. We frankly didn’t know what to expect, from the students or from the AI models. What struck me coming from the students was: first, the enthusiasm to explore; then, as the project developed, they all came to the cool-headed realization of what AI could or could not help them with, and then almost instantly adapted for that," Garnier noted. He expressed confidence that this generation of engineers would avoid simplistic AI reliance, instead maintaining a robust connection with both physical and thought experiments.
A Four-Week Sprint: The Chronology of Innovation
The JARVIS Challenge unfolded as a rapid-fire chronology of design, fabrication, and testing, revealing both the immense power and the inherent limitations of AI in hardware engineering.
Week 1: Initial Designs and AI’s Early Promise
By the end of the first week, one team withdrew, but the remaining six had, with varying degrees of success, developed initial designs for their gas turbines. During this foundational phase, AI tools, particularly LLMs like Claude and ChatGPT, proved highly effective as knowledge amplifiers and organizational aids. Students leveraged AI to summarize complex textbooks on turbomachinery, generate concise explanations of compressible flows, and even learn the intricacies of specialized design software. They tasked AI with sourcing potential vendors for specific components, creating detailed Excel spreadsheets for project management, and answering myriad specific technical questions. One innovative team even created an AI agent within Parley, entrusting it with the role of their project manager, a testament to the students’ eagerness to push the boundaries of AI integration. AI facilitated rapid information retrieval and conceptual brainstorming, allowing teams to quickly overcome initial knowledge gaps.
Week 2: Detailed CAD and Confronting AI’s Limits
The second week marked a critical transition as teams moved from conceptual designs to detailed CAD modeling, component ordering, and combustor prototyping. It was at this stage that the inherent limitations of current AI models began to surface. While AI could offer a multitude of design alternatives and fill knowledge gaps, teams encountered the notorious features of generative AI: hallucinations, sycophancy, and a fundamental lack of physical understanding. These shortcomings often undermined student confidence and introduced errors, paradoxically slowing down the design process.
Elizabeth Tupaj, a member of the victorious 811 Crew, articulated this frustration: "AI is a helpful tool, great at finding information, helping organize things, and can write well, but it can’t do design. The moment the engineer doesn’t know what is going on and the AI is in charge is the moment the design becomes unreliable, at least with AI at its present capabilities." This sentiment was echoed by teaching assistant John Zhang, who observed, "Seeing this firsthand with the students reminded me how much first impressions matter. If the students couldn’t get answers from the AI early on, they quickly grew frustrated and formed a lasting opinion that precluded them from using it later." The challenge highlighted that while AI excels at data synthesis and pattern recognition, it struggles with the nuanced, context-dependent reasoning and deep physical intuition required for complex engineering design.
Final Weeks: Manufacturing Bottlenecks and Human Networks
In the subsequent weeks, as fabrication became paramount, the finalists confronted an obstacle that no AI could resolve: vendor relationships and manufacturing timelines. Students reported that "AI searches found vendors we had no rapport with, who had no interest in our tight timeline." The success in acquiring critical components and meeting tight deadlines often hinged on existing personal relationships and human negotiation skills rather than AI-generated recommendations. This underscored the enduring importance of human networks and communication in the tangible world of hardware development.
First Ignition: Breakthroughs and Lessons Learned
Despite the myriad challenges, the JARVIS Challenge yielded significant breakthroughs. Of the three finalist teams, only "Fast and Fractured" achieved first-attempt ignition of their mini-combustor. This team, despite having no prior gas turbine experience, had heavily leveraged AI for trade studies and architectural comparisons, ultimately arriving at a viable design. The moment of ignition was a powerful validation of the challenge’s core premise.
Masha Folk, the Charles Stark Draper Career Development Professor of Aeronautics and Astronautics, vividly recalled this moment: "The moment that stood out most was when the first student-designed combustor was installed on the test stand. It ignited flawlessly, ramped to full power, transitioned to dual-fuel operation, and then sustained stable combustion on 100 percent Jet-A fuel. This was proof that we can dramatically accelerate the cycle of design, build, and test while giving students hands-on experience with a real engineering challenge." This successful test demonstrated that AI-enabled design, when coupled with motivated students and an ethos of rapid experimentation, could indeed compress traditionally long development cycles.
The Vanguard of AI-Native Engineering: Outcomes and Observations
By the end of May, the two more senior teams — Fast and Fractured and 811 Crew — had completed full engine tests. Fast and Fractured, despite their AI-assisted design prowess, faced persistent delays due to vendor issues. Their hot fire test was unfortunately cut short when the rotor rubbed and seized against the stationary housing, a common yet critical failure point in turbomachinery.
However, Team 811 Crew, which entered the competition with greater exposure to turbomachinery and propulsion concepts, emerged victorious. Their engine successfully started, transitioned to Jet-A fuel, and generated net thrust, fulfilling the challenge’s core objective. PhD student Joe Chiapperi captured the intensity of the moment: "As we stood there with the air-starter, hearing their engines spool up and watching them spit fire, it felt like my heart was racing out of my chest. There were so many ways it could go wrong! What these students accomplished in such a short time span is nothing short of amazing."
Interestingly, the 811 team had shown resistance to heavy AI reliance throughout the competition, preferring to trust their fundamental engineering knowledge and robust teamwork. Tupaj explained their approach: "We had people who were at least somewhat familiar with the design software, mechanical engineers who knew how to build anything, and aerospace engineers who had taken classes on the design of gas turbine engines specifically." This outcome highlighted a crucial tension: while AI offered speed, foundational expertise provided reliability.
A clear observation emerged from the competition regarding AI usage: younger students, particularly first-years, utilized Parley more frequently and creatively, often leaning on it to bridge significant knowledge gaps. Conversely, juniors and seniors, possessing deeper domain experience, were more judicious in their AI application. Professor Andreea Bobu encapsulated this finding: "JARVIS taught me that getting value from AI takes two things: enough expertise to judge what it tells you and catch it when it’s wrong, and enough curiosity to actually lean on it where it could help. The team that moved fastest in the sprint was experienced and leaned heavily on AI to get there. The team that eventually won was more resistant to AI; they had the expertise, but that skepticism made them slower. The sweet spot seems to be knowing enough to stay in charge of the tool, and being eager enough to pick it up in the first place. To me, that’s the real opportunity ahead: training the next generation of engineers who have the judgment to direct these AI tools and the instinct to reach for them."
The competition’s most salient finding was the undeniable role of engineering experience as a multiplier. Human factors, particularly mastering first principles and fundamental concepts, cultivate sound engineering judgment — the ability to navigate complex decisions amidst incomplete information. When it comes to designing and building safety-critical physical systems, human hands and human accountability remain irreplaceable. Teaching assistant Kyle Woody summarized this by stating, "JARVIS has shown that AI copilots can have a multiplicative effect on engineering productivity, with judgment and first-principles thinking serving as the key differentiators among teams."
Broader Implications: Reshaping Engineering Education and Industry
The implications of the JARVIS Challenge for aerospace and other hardware-intensive industries are profound. If small teams, empowered by well-managed AI copilots, can compress design-build-test cycles from years to weeks, the consequences for workforce structure, research and development timelines, and competitive dynamics could be substantial. This paradigm shift could democratize access to complex engineering, enabling smaller entities to innovate at speeds previously reserved for large corporations.
For engineering education, the challenge serves as a powerful testament to the enduring value of foundational knowledge. Professor Cordero, associate director of the MIT Gas Turbine Laboratory, emphasized this point: "JARVIS highlighted the power of AI in the design of physical systems. But it also showed that the key to unlocking that power is education, through coursework, internships, and hands-on extracurriculars like MIT Motorsports and Rocket Team. Performance in JARVIS correlated strongly with year in school. My main takeaway is that in the AI era, education is more valuable than ever." The future engineer will not be replaced by AI but will be augmented by it, requiring a deeper understanding of underlying principles to effectively steer these powerful tools.
The JARVIS Challenge has laid bare the future of engineering: a collaborative ecosystem where human expertise, critical judgment, and advanced AI tools converge. The students who participated in this intense sprint are not merely adapting to a new technological landscape; they are actively shaping it, grappling with the stakes of AI integration not as a theoretical exercise but in the tangible reality of a machine shop, with a live jet engine on the test stand. Their experiences offer invaluable lessons for educators, industry leaders, and policymakers alike, guiding the path toward an "AI-native" engineering future that remains firmly rooted in human ingenuity and responsibility.
