AI at MIT
MIT has become one of the most AI-saturated research universities on the planet, with artificial intelligence now touching nearly every department and lab on campus. Faculty who never planned to work with AI are building machine learning models, training neural networks, and reth...
According to MIT Technology Review's coverage published on April 21, 2026, the transformation happening inside MIT's research buildings is less a planned institutional pivot and more a cascade of individual decisions, unexpected detours, and collegial nudges that have collectively reshaped how science gets done at one of the most influential universities in the world.
Why This Matters
MIT is not just any university. Its research pipelines feed into government policy, defense contracts, pharmaceutical development, and Silicon Valley's next generation of startups. When 90 percent of thesis committees at MIT now involve a significant AI component, as biological engineering professor Angela Koehler told MIT Technology Review, that number signals something seismic about where all of science is heading. Five years ago, AI was a specialty. Today it is infrastructure, and MIT is the clearest proof of that shift.
Daily briefing from 50+ sources. Free, 5-minute read.
The Full Story
Sili Deng did not set out to become an AI researcher. When she joined MIT's mechanical engineering faculty in 2019, her plan was straightforward: build a lab, study combustion kinetics, work on emissions reduction, and investigate flame synthesis of energy materials. Then COVID-19 arrived in 2020 and froze everything. Lab renovations stopped. The physical work she had planned became impossible. Rather than wait, Deng challenged herself and her postdocs to explore machine learning and figure out, given their expertise in combustion science, where the gaps were that AI could fill.
The result of that forced detour was significant. Deng's Energy and Nanotechnology Group developed what researchers call a "digital twin," a computational model that mirrors the real-time behavior of an energy or flow device. Think of it as a living replica of a physical combustion system, one that can eventually predict and control how fuel burns. The practical applications range from cleaner industrial energy systems to more efficient engines. Deng's story is a textbook example of necessity driving innovation in a direction no one originally intended.
Zachary Cordero took a different path. An associate professor of aeronautics and astronautics at MIT, Cordero was not pursuing AI-related research at all until 2024, when John Hart, the head of MIT's Department of Mechanical Engineering, introduced him to Faez Ahmed, an associate professor who specializes in machine learning and optimization for engineering design. The introduction was simple and collegial, but it opened a door. Working with Ahmed and other collaborators on a DARPA-sponsored project, Cordero developed an AI tool that optimizes the material composition of a blisk, which is a bladed disk that serves as a critical component in jet and rocket turbine engines.
Cordero's goal is better engine performance and longer operational life, outcomes that could directly improve reusable rocket engines for heavy-lift launch vehicles. He described the AI system as something that augments human intuition even in situations where human intuition barely exists. That framing is worth sitting with. This is not AI replacing expert judgment but AI extending the reach of that judgment into territory where human pattern recognition simply cannot . These two stories are not outliers. Angela Koehler, the Charles W. and Jennifer C. Johnson Professor of Biological Engineering and faculty lead of MIT HEALS, says she cannot think of a single group meeting where her team is not discussing AI tools. Her lab uses AI models to identify drug candidates that can attach to molecular targets once considered completely off-limits, including transcription factors, RNA-binding proteins, and cytokines. Targets that were labeled "undruggable" are now being approached through computational chemistry powered by machine learning.
Ian Waitz, MIT's vice president for research and the Jerome C. Hunsaker Professor of Aeronautics and Astronautics, put it plainly: "Artificial intelligence is everywhere on campus." He noted that any field dealing with tremendous complexity stands to benefit, listing life sciences, materials science, and image analysis as areas where the AI tools have become standard equipment. Professor Ju Li offered perhaps the most forward-looking observation in the piece, arguing that if AI is given the autonomy to run experiments, fail, iterate, and learn from those failures, it could evolve into something that closely resembles human intelligence.
Key Details
- Sili Deng joined MIT's mechanical engineering faculty in 2019 and pivoted to AI after COVID-19 halted her lab renovations in 2020.
- Deng's Energy and Nanotechnology Group built a "digital twin" capable of predicting and controlling real-time fuel combustion systems.
- Zachary Cordero's AI project was initiated in 2024 after a direct introduction from Department of Mechanical Engineering head John Hart.
- Cordero's work is funded by DARPA and targets performance improvements in blisks used in jet and rocket turbine engines.
- Angela Koehler estimates that 90 percent of the thesis committees she sits on now involve a significant AI component, compared to a much smaller fraction five years ago.
- MIT HEALS, led by Koehler, is using AI to pursue drug targets previously classified as undruggable, including transcription factors and cytokines.
- MIT vice president for research Ian Waitz confirmed AI presence across every major research field at the university as of April 2026.
What's Next
Cordero's DARPA-backed blisk optimization work is aimed at enabling more reliable reusable rocket engines, a milestone that carries direct implications for the commercial space industry and national defense procurement timelines. Deng's digital twin platform, once mature enough to operate in real time, could be adopted by energy companies seeking to reduce emissions without overhauling physical infrastructure. Watch for MIT to formalize more cross-departmental AI collaborations similar to the Hart-to-Ahmed-to-Cordero pipeline, since that model of collegial introduction is already producing funded, publishable research.
How This Compares
Stanford has pursued a similar campus-wide AI integration through its Human-Centered AI Institute, which launched in 2019 and now involves over 200 faculty members across disciplines. But Stanford's approach has been more institutional and top-down, with dedicated funding streams and named centers driving adoption. What MIT's April 2026 coverage reveals is something more organic, where individual researchers are finding their way into AI through pandemic disruptions, hallway conversations, and department head matchmaking rather than through formal programs.
Compare this to what Carnegie Mellon has done with its AI initiative, which has largely concentrated AI talent within its School of Computer Science. MIT's story is different because the researchers making news here, Deng, Cordero, Koehler, are in mechanical engineering, aero-astro, and biological engineering. The AI expertise is spreading outward from computer science into domains where it produces entirely new kinds of research output, not just better algorithms but better rocket engines and better drug candidates.
The broader pattern here aligns with what Google DeepMind demonstrated with AlphaFold in 2020 and its expanded protein structure work in subsequent years: AI is most powerful when it is handed to domain experts who know where the hard problems actually live. MIT's faculty are not becoming computer scientists. They are becoming scientists who use AI, and that distinction is what makes the university's current moment genuinely interesting to watch. Check the AI Agents Daily news feed for ongoing coverage of AI adoption in research institutions.
FAQ
Q: What is a digital twin in AI research? A: A digital twin is a computational model that replicates the behavior of a real physical system. In Sili Deng's work at MIT, the digital twin mirrors a combustion or energy flow device and is designed to predict and control how that system behaves in real time, allowing researchers to test scenarios virtually before applying changes to actual hardware.
Q: How is AI being used to develop new drugs at MIT? A: Angela Koehler's lab at MIT uses AI models to identify molecules that can bind to biological targets previously considered impossible to treat with drugs, including transcription factors and RNA-binding proteins. Machine learning helps identify candidate compounds by analyzing massive datasets of molecular structures far faster than traditional chemistry methods allow.
Q: What is a blisk and why does it matter for rockets? A: A blisk, short for bladed disk, is a single integrated component in jet and rocket turbine engines that combines the rotor disk and blades into one piece. It is critical to engine performance and durability. Zachary Cordero's DARPA-funded AI tool optimizes the material composition of blisks, which could lead to longer-lasting and more reliable engines for reusable heavy-lift launch vehicles.
MIT's story is not about a single breakthrough or a flashy product launch. It is about what happens when one of the world's most research-intensive universities absorbs AI so thoroughly that even professors who never planned to touch it end up building some of its most interesting applications. The next wave of AI-driven discoveries will likely come from exactly these kinds of interdisciplinary collisions, and MIT is running that experiment at scale right now. Subscribe to the AI Agents Daily weekly newsletter for daily updates on AI agents, tools, and automation.
Get stories like this daily
Free briefing. Curated from 50+ sources. 5-minute read every morning.
