Described in a study published Dec. 8 in Nature Electronics, BISC includes a single-chip implant, a wearable “relay station,” and the custom software required to operate the system. “Most implantable systems are built around a canister of electronics that occupies enormous volumes of space inside the body,” says Ken Shepard, Lau Family Professor of Electrical Engineering, professor of biomedical engineering, and professor of neurological sciences at Columbia University, who is one of the senior authors on the work and guided the engineering efforts. “Our implant is a single integrated circuit chip that is so thin that it can slide into the space between the brain and the skull, resting on the brain like a piece of wet tissue paper.” 

Shepard was joined in the BISC effort by senior and co-corresponding author Andreas S. Tolias, PhD, professor at the Byers Eye Institute at Stanford University and co-founding director of the Enigma Project. Tolias’s pioneering work training AI models on large-scale neural datasets — including datasets recorded in the Tolias laboratory using BISC — enabled the team to evaluate the device’s neural decoding performance. “BISC turns the cortical surface into an effective portal, delivering high-bandwidth, minimally invasive read–write communication with AI and external devices,” Tolias says. “Its single-chip scalability paves the way for adaptive neuroprosthetics and brain-AI interfaces to treat many neuropsychiatric disorders, such as epilepsy.”

Dr. Brett Youngerman, assistant professor of neurological surgery at Columbia University and a neurosurgeon at NewYork-Presbyterian/Columbia University Irving Medical Center, served as the chief clinical collaborator on the project. “This high-resolution, high-data-throughput device has the potential to revolutionize the management of neurological conditions from epilepsy to paralysis,” he says. Youngerman, Shepard, and NewYork-Presbyterian/Columbia epilepsy neurologist Dr. Catherine Schevon were recently awarded a grant from the National Institutes of Health to implement BISC in the management of drug-resistant epilepsy. “The key to effective brain-computer interface devices is to maximize the information flow to and from the brain, while making the device as minimally invasive in its surgical implantation as possible. BISC surpasses previous technology on both fronts,” continues Youngerman.

“Semiconductor technology has made this possible, allowing the computing power of room-sized computers to now fit in your pocket,” Shepard says. “We are now doing the same for medical implantables, allowing complex electronics to exist in the body while taking up almost no space.”

Smaller, Safer, and Faster

BCIs work by interfacing with the electrical signals that neurons use to transfer information throughout the brain. Today’s state-of-the-art BCIs, used in medical contexts, are constructed from individual microelectronic components, including amplifiers, data converters, radio transmitters, and power management circuits. To accommodate all these devices, a large canister of electronics must be surgically implanted in the body, either by removing a portion of the skull or by placing the device in another location, such as the chest, and running wires to the brain.

BISC works differently. The entire implant, which occupies less than 1/1000th the size of a conventional device, is a single complementary metal-oxide-semiconductor (CMOS) integrated circuit chip thinned to just 50 μm. With a total volume of approximately 3 mm³, the flexible chip conforms to the surface of the brain. This micro-electrocorticography (µECoG) device integrates 65,536 electrodes, 1,024 simultaneous recording channels, and 16,384 stimulation channels. By leveraging the large-scale manufacturing techniques developed in the semiconductor industry, these implants can be easily manufactured at scale.

The single-chip implant includes a radio transceiver, wireless powering circuit, digital control, power management, data conversion, and the analog circuits required to support the recording and stimulation interfaces. The battery-powered relay station powers and communicates with the implant, transferring data via a custom ultrawideband radio link that achieves 100 Mbps data bandwidths — a connection with at least 100 times higher throughput than any competing wireless BCI device. The relay station is itself an 802.11 WiFi device, in effect forming a relayed wireless network connection from any computer to the brain. 

BISC has its own instruction set, supported by an extensive software stack, which together constitute a computing architecture designed for BCIs. As demonstrated in this study, these high-bandwidth recording capabilities allow brain-signal patterns to be submitted to advanced machine-learning or deep-learning frameworks for decoding complex intentions, perceptions, or states.

 “By integrating everything on one piece of silicon, we’ve shown how brain interfaces can become smaller, safer, and dramatically more powerful,” Shepard says.

The BISC implant was manufactured using TSMC’s versatile 0.13-μm Bipolar-CMOS-DMOS (BCD) technology. This manufacturing process integrates three technologies onto a single chip to create mixed-signal integrated circuits (ICs). This integration enables the efficient combination of digital logic (from CMOS), high-current and high-voltage analog functions (from bipolar and DMOS transistors), and power devices (from DMOS), all of which are essential for BISC.

From Lab to Clinic

To make this technology available to doctors and patients, Shepard’s group partnered closely with Youngerman at NewYork-Presbyterian/Columbia University Irving Medical Center. Together, they refined surgical methods to safely implant the paper-thin device in a preclinical model and demonstrated its recording quality and stability, as described in the current study. Studies in human patients for short-term intraoperative recordings are underway.

“These initial studies give us invaluable data about how the device performs in a real surgical setting,” Youngerman says. “The implants can be inserted through a minimally invasive incision in the skull and slid directly onto the surface of the brain in the subdural space. The paper-thin form factor and lack of brain-penetrating electrodes or wires tethering the implant to the skull minimize tissue reactivity and signal degradation over time.”

Extensive pre-clinical testing of BISC in the motor and visual cortices drew on collaborations with both Dr. Tolias and Bijan Pesaran, professor of neurosurgery at the University of Pennsylvania, both of whom are leaders in computational and systems neuroscience.

“The extreme miniaturization by BISC is very exciting as a platform for new generations of implantable technologies that also interface with the brain with other modalities such as light and sound,” Pesaran says.

Developed under the Neural Engineering System Design program of the Defense Advanced Research Projects Agency (DARPA), BISC combines Columbia’s strengths in microelectronics, Stanford’s and Penn’s cutting-edge neuroscience, and NewYork-Presbyterian/Columbia University Irving Medical Center’s surgical innovation.

Toward Real-World Applications

To accelerate translation, the Columbia and Stanford teams launched Kampto Neurotech, a spin-off company founded by Columbia electrical engineering alumnus Dr. Nanyu Zeng, one of the project’s lead engineers. Kampto Neurotech is developing commercial versions of the chip for preclinical research applications and raising funds to advance the system toward human use.

“This is a fundamentally different way of building BCI devices,” Zeng says. “In this way, BISC has technological capabilities that exceed those of competing devices by many orders of magnitude.” 

In a technological landscape driven by advances in artificial intelligence, BCI technologies have drawn considerable recent interest in both restoring function to those affected by neurological conditions and in potentially augmenting human capabilities by providing direct interfaces to the brain.

“By combining ultra-high resolution neural recording with fully wireless operation, and pairing that with advanced decoding and stimulation algorithms, we are moving toward a future where the brain and AI systems can interact seamlessly — not just for research, but for human benefit,” says Shepard. “This could change how we treat brain disorders, how we interface with machines, and ultimately how humans engage with AI.”

Lead Photo Caption: The BISC implant shown here is roughly as thick as a human hair. 

Lead Photo Credit: Columbia Engineering

About The Study

Journal: Nature Electronics

DOI: 10.1038/s41928-025-01509-9

Title: Stable, chronic in-vivo recordings from a fully wireless subdural-contained 65,536-electrode brain-computer interface device

Authors: Taesung Jung, Nanyu Zeng, Jason D. Fabbri, Guy Eichler, Zhe Li, Erfan Zabeh, Anup Das, Konstantin Willeke, Katie E. Wingel, Agrita Dubey, Rizwan Huq, Mohit Sharma, Yaoxing Hu, Girish Ramakrishnan, Kevin Tien, Paolo Mantovani, Abhinav Parihar, Heyu Yin, Denise Oswalt, Alexander Misdorp, Ilke Uguz, Tori Shinn, Gabrielle J. Rodriguez, Cate Nealley, Sophia Sanborn, Ian Gonzales, Michael Roukes, Jeffrey Knecht, Daniel Yoshor, Peter Canoll, Eleonora Spinazzi, Luca P. Carloni, Bijan Pesaran, Saumil Patel, Joshua Jacobs, Brett Youngerman, R. James Cotton, Andreas Tolias, Kenneth L. Shepard

Funding/Acknowledgments: This work was partly supported by the Defense Advanced Research Projects Agency (DARPA) under Contract N66001-17-C-4001, the Department of the Defense Congressionally Directed Medical Research Program under Contract HT9425-23-1-0758, the National Science Foundation under Grant 1546296, and the National Institutes of Health under Grant R01DC01949

The final panel, moderated by Bayan Bruss of Capital One, focused on the “Future of AI and Fintech” and featured academic researchers from Columbia Engineering, the University of Texas at Austin, and Brown University. Their conversation explored the integration of quantum computing and AI into financial services, including the current limitations and potential of these technologies.

“In terms of where we’re at with quantum computing, things are very early-stage,” said Henry Yuen, Srivani Family Associate Professor of Computer Science at Columbia Engineering. “There’s a lot of fundamental research to be done. If we build these machines, what are we going to use them for?”

The panel agreed that significant theoretical and practical advancements are needed before it can revolutionize finance, particularly in the areas of scalability and reliability. 

“We had a fantastic lineup of speakers,” said Capponi, closing out the day’s agenda. “It’s a very good example of the mix between academia and industry. We all learn from each other. I’ve seen interactive discussions with people that hopefully will lead to further ideas.” 

Prem Natarajan, chief scientist and head of Enterprise AI at Capital One, also praised the breadth of expertise among the participants.  

“We value gatherings like this to foster discussion and potential collaborative ideas,” said Natarajan. “Successfully maximizing the broad benefits of AI  will require interdisciplinary collaboration, combining expertise from academia's research capabilities with industry's infrastructure and data resources. The breakthroughs of the future will come out of such partnerships.”

Lead Photo Caption: From left to right: Bayan Bruss, VP, Applied AI Research at Capital One, Atlas Wang, associate professor of electrical and computer engineering, UT Austin, Henry Yuen, Srivani Family Associate Professor of Computer Science, Columbia Engineering, Randall Balestriero, assistant professor of computer science, Brown University

During the opening keynote presentation, Carnegie Mellon’s Priya Narasimhan shared that she didn’t set out to build a global sports technology company. But one night in 2008 at a Pittsburgh Penguins game, watching from the nosebleed section, she realized how much of the action she was missing—angles, stats, insights that could make the experience richer. That moment led to a research project with her computer science students and ultimately to the launch of her startup YinzCam.

YinzCam started as an experiment to deliver real-time replays and stats directly to phones and has grown into a platform serving more than 160 professional teams around the world. Today, it powers apps, websites, AR features, and loyalty programs, continuing with its goal to keep sports fans connected beyond game day.

For Narasimhan, the fan’s perspective has always been the starting point. “At its core, the company is about personalization—turning data into experiences that deepen loyalty and engagement.”

And while the data now says 12 percent of fans actually experience games in a stadium, the bottom line is still the same: from broadcasts to merchandise, technology isn’t the end goal; the fan is. As Saj Cherian, chief of staff and head of Fanatics Ventures, explained, “If you’re in the consumer business, your relationship to your customer, the sports fan, is so important.” Trust isn’t just a box to check, he added,  “Trust … is a brand asset.” Something that takes years to build and moments to lose. 

“When you get into these numbers, you get into these analytics, it can get really impersonal pretty quickly,” said Julie Souza, global head of sports at Amazon Web Services. “I get people coming at me all the time, saying the numbers have ruined sports. If that’s what it’s coming across as—if it is alienating—we are doing it wrong.”

The takeaway? Sports is a business, they said, but emphasized its heart is people. Profits only follow when fans feel seen, respected, and genuinely part of the game.

But it isn’t just the technology that matters—it’s the people who build it. Behind every new tool or application are researchers, engineers, and students shaping how AI is applied to sports. In the Data Science & Analytics Across Sporting Ecosystems panel, speakers turned the spotlight on the next generation, offering practical advice on how students can prepare for a rapidly evolving work and technology landscape.

The advice from the panel was clear: Don’t limit yourself. “There are only so many teams that you can work for,” said Konstantinos Pelechrinis, professor of informatics & network systems at the University of Pittsburgh and a data analytics consultant for the Dallas Mavericks, “but there are so many other companies that work with these teams that can give you a foot in the door doing pretty interesting things with AI.” Instead of focusing narrowly on a league or franchise, he said students should think about the wider ecosystem of industries, data platforms, startups, and vendors that are shaping the field.

Equally important is skill-building. “I would focus on building skills rather than focusing on a really specific field,” Ozan Adiguzel, principal AI engineer at Finster AI, advised, stressing that coding, system design, and statistics are foundational across industries. 

While tools like Python, TypeScript, and even emerging languages such as Rust or Golang open doors, understanding the mathematics behind algorithms matters just as much. 

“Having a good foundation in math and statistics is important,” added Jorge Ortiz, associate professor of electrical and computer engineering at Rutgers University and CV/AI lead for the New York Yankees. “There's still verification and making sure that your code is working correctly.” 

Students who build strong skills in math, statistics, and coding — and demonstrate them through class projects or public portfolios — will find opportunities not only in sports but across industries that are eager for computer scientists.

The lightning talks from Columbia Engineering researchers highlighted cutting-edge research at the intersection of sports, health, technology, and human behavior. Computer Science Assistant Professor Yunzhu Li showed how hybrid modeling—combining physics-based equations with machine learning—can help robots and AI systems adapt to the unpredictable dynamics of sports. Parth Gami, a 5th-year biomedical engineering PhD student, introduced a noninvasive ultrasound method for tracking cardiovascular health in athletes, with potential for wearable fitness applications. Mechanical Engineering Professor Sunil Agrawal shared how his lab uses robotics to study and train human movement, from everyday actions to competitive sports, with promising results for improving performance and preventing injury. And finally, Assistant Professor Lydia Chilton’s talk on digital twins showed how large language models can simulate and anticipate human behavior at scale—whether in shopping, education, or event planning—with sports as the next frontier.

“As much as we are advancing algorithms and data models, the essence of this work is human,” said Eddie Mandhry, managing director at Columbia–Dream Sports AI Innovation Center. “Whether it’s athletes pushing their limits, fans connecting with the game, or researchers and engineers designing new tools, AI in sports is ultimately about enhancing human potential and experience.”

Lead Photo Caption: Keynote speaker Priya Narasimhan, professor at Carnegie Mellon and CEO at YinzCam, talks with Vishal Misra, vice dean of computing and AI at Columbia Engineering

Lead Photo Credit: Eileen Barroso