MIT EECS Department

09/02/2026

Gohar Chaudhry, a graduate student in the MIT EECS Department, identified a problem at the heart of modern AI systems. Agentic workflows—complex systems that chain together multiple AI models and tools to tackle complicated tasks—are getting very inefficient. Developers have to hard-code every technical choice upfront, from which models to use to which hardware to run them on. It's nearly impossible to do optimally.

So Chaudhry and his team at MIT and Microsoft developed Murakkab, a system that does the hard work automatically. Developers describe what they want the workflow to do in plain language. Murakkab figures out the best models, tools, and hardware configuration. It even adjusts those configurations in real time based on whether the user prioritizes speed or cost.

"Energy usage is a huge concern, so we need to be very careful about how efficient these workflows are," Chaudhry says. "It is very easy to over-allocate resources, wasting energy and money." When tested on video Q&A and code generation tasks, Murakkab used only 35 percent of the computation of traditional approaches—consuming 27 percent as much energy for 25 percent of the cost.

https://www.eecs.mit.edu/improving-the-speed-and-energy-efficiency-of-ai-agents/

09/02/2026

“My original plan was to be at MIT for a couple of years and then go back to France, but 27 years later, I’m still here,” explains Fredo Durand, the Amar Bose Professor of EECS. What keeps him here? The "abundance of smart people in every field", which has allowed Durand to tackle problems ranging from bridge vibration to medical recordkeeping to the limits of color photography.

MIT professor of computing Fredo Durand explores the limitless possibilities for synthetic images, computational photography

08/31/2026

Retail workers spend roughly 50 percent of their time managing inventory—a $15 billion problem in the U.S. alone. Cartesian, co-founded by Associate Professor Fadel Adib in the MIT EECS Department and Isaac Perper, is solving it using technology invented at MIT.

The platform uses wireless signals from RFID tags to locate products in real-time across stockrooms and shop floors. Cartesian is already deployed in more than 700 stores across 15 countries, working with major retailers including Inditex (ZARA, Pull&Bear, Oysho).

"The big problem we're solving is that about 50 percent of working hours in retail stores go to managing inventory," says Adib. "We use algorithms to decipher indoor locations using wireless signals."

Beyond retail, the technology has applications in manufacturing, logistics, and robotics. Cartesian's team aims to expand to tens of thousands of stores over the next year before moving into other industries.

https://www.eecs.mit.edu/startup-helps-retailers-track-their-products-in-real-time/

08/28/2026

Chipmakers and operating system developers have spent years building defenses. A new study from the lab of Mengjia Yan shows that a key assumption behind many of them doesn’t hold.

MIT CSAIL researchers found a way to exploit a split-second gap in chip security — and used it to acquire a Linux system’s password file.

08/28/2026

MIT researchers led by Vivienne Sze, a professor in the MIT EECS Department, developed a chip called Gleanmer that constructs detailed 3D maps using only about 6 milliwatts of power—roughly what a single LED uses.

That changes what's possible. A tiny drone could zip through an industrial HVAC system, checking for gas leaks. AR glasses could run all day without draining the battery. Any battery-limited device can understand its environment in real time—something that previously required power-hungry systems and massive memory.

The breakthrough came from rethinking how robots represent space. Instead of storing rigid 3D pixels, the system uses flexible ellipsoid shapes that adapt to curved objects more efficiently. Then the team designed specialized hardware to accelerate that algorithm, keeping data moving through fast on-chip memory instead of power-hungry storage.

"Real-time 3D mapping has been the missing piece for small autonomous systems," says Sertac Karaman, a professor in MIT Aeronautics and Astronautics and co-author. "Gleanmer makes that possible for the first time in a chip you can hold between your fingers."

https://news.mit.edu/2026/new-chip-could-help-tiny-robots-traverse-complex-environments-0623

08/19/2026

AI can do lots of things--comb through voluminous data, identify patterns, flag changes--but it can't fold your socks. Why not? Associate Professor Vincent Sitzmann explains why mundane tasks in the physical world are still so challenging for this technology.

07/15/2026

As a software engineer at the prestigious New York art auction house, Sotheby’s, Kelly Shen ’17 works in the growing field of art intelligence. She builds algorithms to predict artwork prices using factors like an artist’s name, popularity, and exhibitions, as well as art buying trends and historical data. Her work has included everything from cataloging systems to early efforts around crypto transactions for high-value sales.

Shen’s role combines her technical academic background, as a double major in the MIT EECS Department and math, and her passion for art, fueled by her creative mother who encouraged her love of drawing at an early age. The lessons she learned at MIT have also stuck with her, notably the appreciation she gained for blending innovation with precision.

“I love the challenges of considering feasibility and our legacy systems and coming up with the best middle ground,” Shen says. “You can build a super-sophisticated algorithm, but if it doesn't bring more audience engagement, it doesn’t really matter.”

Photo courtesy of Kelly Shen.

https://alum.mit.edu/slice/art-and-algorithms-sothebys

07/13/2026

Lecturer Gim Hom recently taught a "last class" in the Cypress Engineering Design Studio here at EECS. His students were a special group: fifteen of his classmates from the class of 1971.

A recent Reunion activity, taught by EECS Senior Lecturer Gim Hom, doubled as both his “last class” and a chance to reconnect with classmates.

07/10/2026

A new study from MIT researchers could help to enable next-generation lidar sensors — which use light in the form of a pulsed laser to measure variable distances to the Earth — that are compact, durable, and have no moving parts.

This new demonstration could fuel the development of advanced lidar sensors for demanding applications like autonomous vehicle navigation, aerial surveying, and construction site monitoring.

“The functionality we demonstrated in this work solves a fundamental problem for integrated optical-phased-array technology, enabling future lidar sensors that can achieve significantly higher performance than we could demonstrate previously,” says MIT EECS Department Professor Jelena Notaros, senior author of a paper on this innovation which appears in Nature Communications.

https://news.mit.edu/2026/photonics-advance-could-enable-compact-high-performance-lidar-sensors-0507