Geo Tu earned his B.S. in electrical engineering from Texas ECE in 2008. He went on to earn an M.S. in electrical engineering from the University of Southern California in 2010. Now an architect at Lightmatter, Tu is helping develop technology designed to address one of computing’s most pressing challenges: moving enormous volumes of data quickly and efficiently.
At Lightmatter, Tu collaborates across teams to consolidate technical requirements, assess feasibility and define architecture specifications for the company’s Passage product line. His work ranges from specifying chip- and system-level requirements including control hardware for stabilizing photonic circuits to developing link-training algorithms.
His path to Lightmatter was built as much through relationships as technical expertise.
“The semiconductor industry is truly a small world, and my story is an example of how important it is to maintain your professional and personal networks,” Tu said.
An overlap at UT Austin with Darius Bunandar first put Lightmatter on Tu’s radar. Bunandar later reached out as a potential customer of Analog Devices during Lightmatter’s early days. Ultimately, however, it was an old friend and former colleague who made the strongest case for joining the startup.
With no previous background in photonics, Tu also turned to friends in the field, including fellow Longhorns, to better understand the underlying technology. That willingness to learn across disciplines has been a recurring theme throughout his education and career.
Finding a Home at UT
Growing up in Houston as an “elder Millennial,” Tu became fascinated by computers and technology as the industry rapidly expanded. By high school, he was considering electrical engineering, medicine or a career combining the two.
UT Austin stood out for its highly regarded engineering programs and the freedom it offered students to explore courses across different colleges. Its extensive alumni and industry network also promised access to internships and full-time opportunities.
Several close high school friends, including his future college roommates, also chose UT. At the same time, the university was large enough to provide nearly limitless opportunities to meet new people. Some of the most meaningful connections Tu formed at UT extended far beyond the classroom. During his freshman year, he met his future wife, Heidi, while the two lived in dorms across the street from one another.
“The best memories of my time at UT were dating my wife,” Tu said.
Their connection to the university has endured. To recognize the value of their education and help future students pursue similar opportunities, the couple established an endowed scholarship for engineering students.
Technical Achievement and Team Trust
Among Tu’s proudest professional achievements was leading the digital design of a market-defining successive-approximation-register analog-to-digital converter, commonly known as a SAR ADC.
The product achieved first-silicon success, established an industry-standard performance benchmark that still holds, was presented at the IEEE International Solid-State Circuits Conference and received an innovation award at Analog Devices’ General Technical Conference.
Yet Tu does not measure professional pride solely through successful products, publications or awards. He finds equal satisfaction in becoming someone his colleagues can depend on.
“Another recurring proud moment is reaching the point on a team where your colleagues half-joke about wanting to clone you,” Tu said. “Building that level of trust, reliability and camaraderie with the people you work with every day can be just as rewarding as any tapeout.”
That perspective reflects the three types of people who have exerted the greatest influence throughout Tu’s academic and professional life: visionaries, true believers and creative engineers.
Visionaries helped him identify major industry shifts and recognize high-impact opportunities. True believers demonstrated how infectious enthusiasm can turn technical work into genuine professional fulfillment. Creative engineers modeled a balance between respecting conventional wisdom and challenging it, using first principles to develop solutions that go beyond textbook approaches.
The Future of ECE
Looking ahead, Tu sees electrical and computer engineering moving rapidly toward heterogeneous 2.5D and 3D integration, chiplets and advanced packaging. As computing systems grow more complex, overcoming physical interconnect bottlenecks has become a defining challenge.
That challenge is driving the development of high-density silicon photonics and optical input/output technologies positioned directly at the package boundary. At the same time, computing architectures are increasingly being developed alongside the software models they are intended to run.
“This shift, combined with interconnect constraints, is opening lanes for agile, specialized startups to co-design and integrate directly into tier-one supply chains,” Tu said.
The most consequential opportunities, he believes, often exist at the boundaries between different hardware domains and between silicon and software. Engineering education must therefore continue to balance deep, discipline-specific knowledge with a broader understanding of system architecture.
“Developing fluency across key complementary domains differentiates good engineers from great ones,” Tu said.
Universities must also prepare students to use artificial intelligence as a fundamental engineering tool. AI-assisted architectural reasoning, design-tradeoff analysis and verification should become core parts of the curriculum and its assessment. Just as importantly, undergraduate students should receive direct exposure to modern design workflows, tapeouts and post-silicon validation.
Advice for the Next Generation
For students beginning their studies, Tu recommends taking full advantage of UT’s flat-rate tuition by enrolling in courses simply out of curiosity. Some of his favorite memories came from engineering courses outside his primary discipline, as well as physical education classes such as ballroom dancing, volleyball and weightlifting.
“Stepping outside your major and college doesn’t just widen your perspective, it’s one of the best ways to build lasting friendships, and it teaches you how to connect across diverse circles throughout life,” Tu said.
Students should also practice with AI tools regularly, both to reinforce foundational knowledge and to develop workflows they can use professionally. The technology’s effectiveness, however, remains tied to the judgment and expertise of the person using it.
“The earlier you learn where it accelerates thinking versus where it falls short, the larger your competitive edge,” Tu said.
For students preparing to graduate, compensation deserves careful attention. Starting salary affects more than the first year of income; it also establishes the baseline for future raises, promotions and retirement contributions.
“Don’t shy away from interviewing broadly to discover your market value, and always negotiate your offers,” Tu advised. “Great companies recognize high-potential talent and expect to compete for it.”
Above all, Tu encourages early-career engineers to remain intensely curious. They should seek additional responsibilities, learn how adjacent teams operate and build relationships beyond their immediate groups. They should also work to understand how their individual contributions fit into a larger product and how that product succeeds in the marketplace.
“That systems-level perspective,” Tu said, “is what separates junior contributors from true leaders.”
Building Lightmatter’s Future
Tu’s connection to Lightmatter can be traced in part to fellow Longhorn Darius Bunandar, a Lightmatter co-founder and the company’s chief scientist. Bunandar earned his B.S. in mechanical engineering from UT Austin in 2013 before completing his doctorate at MIT, where he studied quantum computing with nanophotonics.
Today, Bunandar leads research and development across Lightmatter’s device physics, integration and architecture, work that encompasses several generations of the company’s technology.
“Lightmatter builds photonic chips that move data between the GPUs that run AI,” Bunandar said. “I lead the company’s research and development: the device physics, the integration and the architecture that span multiple product generations.”
His days are divided among technical investigation, team leadership and long-range product planning. On the technical side, Bunandar works with device, packaging and test engineers to analyze measurements from test chips and determine what the results mean. As a manager, he focuses on hiring, setting priorities and giving the people closest to each problem the freedom to solve it.
The third part of his role involves looking several generations ahead.
“It means deciding which bets we make two or three generations out, and defending those choices with our designers, our manufacturing partners and our customers,” he said.
From MIT Research to a New Company
The idea behind Lightmatter emerged while Bunandar was pursuing his doctorate at MIT. He and co-founder Nick Harris overlapped in the university’s photonics community and recognized an emerging problem: Even in machine learning’s early days, demand for computing power was increasing at a pace that conventional transistor scaling would not be able to sustain.
They also saw a potential solution. For some of the essential work computers perform, light can move information more effectively than copper.
“The demand for compute was growing on an exponential that transistor scaling would not satisfy, and light is simply a better medium than copper for some of the work a computer has to do,” Bunandar said.
Bunandar, Harris and Thomas Graham founded Lightmatter in 2017. They tested their concept through the MIT $100K Entrepreneurship Competition, which they won, and raised the company’s first round of funding soon afterward.
The early insight behind the company has become increasingly relevant as artificial intelligence models have grown larger and more computationally demanding. Lightmatter’s technology uses photonics to address the challenge of moving data among the graphics processing units that power those models.
Preparing Students for a Rapidly Changing Industry
As technological development accelerates, universities face the difficult task of preparing students for an industry they may not enter for another four or five years. Bunandar believes the solution is not to build an education around whichever tools are most popular at a given moment. Instead, students need a strong command of enduring fundamentals and the ability to learn unfamiliar fields quickly.
“With the exponential improvement in computing, the specific tools and model architectures that are hot today have a half-life of maybe 18 to 24 months,” he said. “Fundamentals like Maxwell’s equations, semiconductor physics, heat transfer and linear algebra have no half-life at all.”
The engineers who adapt most successfully, he said, combine that foundational knowledge with a demonstrated ability to teach themselves something new.
“Every engineer I’ve hired who ramped up fast had deep fundamentals and had proven at least once that they could teach themselves a new field in a few months,” Bunandar said.
Universities can help students cultivate that skill by encouraging them to explore the boundaries between academic disciplines. Those boundaries are increasingly where industry’s most difficult and least neatly categorized problems can be found.
Bunandar’s own education illustrates the value of crossing those traditional divisions. He left UT with backgrounds in both physics and mechanical engineering, a combination that has proved particularly relevant in photonics.
“In photonics, thermal gradients, warpage, package cooling and the mechanics of attaching hundreds of optical fibers are some of the most challenging problems in the industry, not the photonic device physics themselves,” he said.
Seek Out the Difficult Problems
Bunandar’s advice to undergraduate engineering students begins with a straightforward challenge: Do not avoid the difficult courses or problems.
“Take the difficult classes, grind and work through it,” he said. “Nobody at Lightmatter has ever asked a candidate about their GPA, but we constantly ask about the one extremely difficult technical challenge that they’ve overcome.”
He also encourages students to become fluent in a second discipline. Many of the strongest engineers he works with can operate comfortably in two complementary fields, such as photonics and packaging, computer architecture and compilers, or mathematics and analog devices.
“The best engineers I work with are bilingual,” Bunandar said. “The boundary is where the breakthroughs are.”
Technical knowledge alone, however, is not enough. Engineers must also be capable communicators who can explain the value of their work and earn support for promising ideas.
“Learn to write and speak clearly,” Bunandar said. “You don’t have to sell yourself, but you do have to sell your ideas.”
Have a suggestion for an alum to spotlight in the future? Contact Colton Lathram.
