Audrey Woods, MIT CSAIL Alliances | January 20, 2026
As the productive momentum of Moore’s Law fades, specialized hardware is stepping in to fuel the next wave of computing technology. Accelerators, compilers, and programming languages are now critical to provide the kind of progress that computer scientists once took for granted, and the hungry demands of AI will only increase that need.
MIT CSAIL Assistant Professor Rachit Nigam—who joined the lab in January, 2026—works at the intersection of programming languages and computer architecture, with the goal of creating solutions that work across the programming stack to support the creation and deployment of specialized hardware. Balancing design, verification, and efficiency, Professor Nigam is excited to bring clarity to a complex field and offer improvements that have real-world impact.
FINDING HIS INTEREST: HIGH SCHOOL CODING & PHD AWARDS
While Professor Nigam was introduced to various programming languages at school—Visual Basic, HTML, Logo—the first time he fell in love with coding was when he was presented with the “utterly inaccessible” blue background and white font of the Borland C programming language environment. After that, he spent most of his high school writing games and making simple AI agents to play them. This passion brought him to the University of Massachusetts Amherst to study computer science. There, Professor Nigam was introduced to the concept of functional programming, a different style of writing code which treats functions (little blocks of code that do something) as the main building blocks rather than the step-by-step instructions typical of most programming approaches. “Functional programming changed the way I understood programming. After that, I emailed half the department at UMass trying to figure out who can teach me more about this.” Professor Nigam quickly became involved with programming languages research.
His interest in computer architecture research came when he started at Cornell University. Broadly wanting to use programming languages techniques to build systems, he joined his PhD advisor’s lab and started exploring how people design hardware. He found the duality of programming languages—which describe what computation to perform—and computer architecture—which decide how to perform computations—particularly interesting. His PhD thesis, “Modular Abstractions for Efficient Hardware Design,” captured a fundamental tension in hardware design: the need to express and reason about time in a way that software programs do not. His dissertation won the John C. Reynolds Doctoral Dissertation Award from the ACM Special Interest Group on Programming Languages (SIGPLAN), which highlighted the “groundbreaking rethinking of hardware design… enabling predictable and composable hardware semantics.” It was also an honorable mention for the Outstanding Dissertation Award for the Special Interest Group on Computer Architecture (SIGARCH) and the IEEE Computer Society’s Technical Committee on Computer Architecture (TCAA).
Now, after a year at Jane Street (a CSAIL Alliances Affiliate) implementing his work in industry, Professor Nigam leads the Foundations of Languages and Machines (FLAME) Lab at CSAIL where he’s working on better ways to create specialized hardware and support the future of computing technology.
RESEARCH: THREE PILLARS OF HARDWARE DESIGN
Professor Nigam explains how his research must balance three major considerations: productivity, efficiency, and correctness. Before Moore’s Law and Dennard scaling ended, there was a “neat boundary between the hardware and the software.” Companies like Intel, AMD, and NVIDIA designed computer chips and programmers built software on top of them. With the end of process scaling, software companies like Google, Meta, and Amazon have increasingly invested in designing custom chips that can accelerate the specific computations they care about. These chips are not a one-time investment: when computational demands change, new chips with completely different circuitry are deployed. Professor Nigam’s research breaks down the challenges of rapidly building such chips into three questions, “How do you quickly describe what your chip does (productivity)? How do you make sure that it is an efficient circuit (efficiency)? And how do you make sure it actually does what you want it to do (correctness)?” In an ideal world, there would be a common programming language to turn high-level computations into efficient circuits with formally guaranteed correctness. But “most existing programming systems make a tradeoff between the three of these. Finding a solution that addresses all three questions in a way that people’s ability to build computer chips is dramatically improved is a really exciting research question.”
One solution he’s developed is Filament, a hardware description language designed to automatically ensure hardware pipelines are composed correctly. Building on the industry standard language Verilog, Professor Nigam explains how Filament added a timing component which enables modularity and generates error messages rather than just allowing the design to silently break. He believes one reason his thesis and research have resonated so strongly in the computer science community is because they asked the question: “how important is time in defining modularity?” Previous methods tended to abstract away time, seeing the problem as circuits and behaviors. But Professor Nigam has identified that to build “really good hardware, you have to think about time.” Filament—as well as some of the other systems Professor Nigam has built—aimed to identify exactly what role time played. Furthermore, Filament doesn’t compromise on efficiency and offers formal correctness guarantees. Filament has already had an impact in industry via Professor Nigam’s relationship with Jane Street, who approached him during his PhD because they wanted to build hardware and deploy it quickly without compromising on verification.
Another tool Professor Nigam has created is Calyx, a system for automatically transforming high level programs into circuits. “Think of taking languages like C or Python, which are what normal programmers work with, and automatically turning them into efficient circuits. Calyx is the compiler infrastructure that enables this.” Compilers, basically, turn one programming language into another language for a given situation. “For example, when you want to run a program on your CPU, your CPU understands something called Assembly, which is a very low-level way of defining programs. What a compiler will do is take a language like C where you can define things like conditionals and loops, and translate that into Assembly.” Calyx lets users take programs written in familiar, high-level languages (like Python or C-style code) and automatically turn them into efficient hardware designs, giving them a way to build custom accelerators without needing to be chip experts. Excitingly, Calyx is integrated with the LLVM CIRCT Project, an umbrella project to build open-source hardware design tools, which Professor Nigam calls a “big achievement.”
“When you do research, there’s work that will get you the next paper or intellectual idea and then there’s work that helps your users. For large-scale systems like Calyx, there’s a lot of bug fixing, helping users, writing documentation, building tools, writing error messages, etc. This doesn’t directly lead to research, but personally, it is still valuable because building and maintaining a system gives you a lot of expertise and experience that allows you to tackle much deeper research problems.” Finding a balance between publishing big ideas and getting tools into peoples’ hands in the fast-moving environment of industry is something that motivates Professor Nigam’s research and inspires him to keep thinking deeply about problems while working with companies and their real-world challenges.
LOOKING AHEAD: CHALLENGES IN DESIGN AND USE OF SPECIALIZED HARDWARE
There is a big tension between what Professor Nigam calls “programming tools for the masses” (languages like Python or JavaScript) and “programming tools for FLOPS” (floating-point operations per second). “A vast majority of the world’s compute capacity (FLOPS) lives within chips like GPUs and specialized accelerators, and programming them looks very different from writing software programs. You have to carefully utilize the resources over time to get maximum performance.” He sees an increasing interest in, and need for, new programming models that can make using such specialized chips easier. “In some ways, the challenges of designing and using hardware are the same: they both need you to think about the effect of resources and time.” Developing both the theoretical foundations to answer such questions, and systems that have real-world impact drives Professor Nigam’s research forward.
Similarly, Professor Nigam believes verification will become more important going forward. “As we build larger-scale systems, it becomes harder to figure out how the interactions change the behavior of our system. What I would love to do is figure out how to help people build the right thing quickly without compromising on efficiency or design.” Fundamentally, Professor Nigam believes “we’re building tools when we build computers or computer systems. The tools had better be right.” Such tools are even more important in the agentic AI era where “AI systems increasingly rely on tools to get feedback on programs they’re generating or debugging. Building formally grounded tools will enable these AI systems to rapidly iterate and design new systems.”
As more industry players begin to build specialized hardware, Professor Nigam encourages companies to “make big bets on building better tooling that allows engineers to collaborate across the hardware–software boundary and give them the automation to iterate quickly.” He emphasizes the opportunities available to those who collaborate with academia, “because designing compilers or programming languages in these specialized domains is not a solved problem. We need a lot more collaboration to distill and unify things in the same way programming paradigms in the software community have.”
Motivated both by the joy of working with students and the intellectual challenge of coming up with elegant solutions to complex problems, Professor Nigam is excited to continue his work at MIT CSAIL and join a community of deep-thinking researchers advancing the state of computer technology: “Alan Kay said, ‘People serious about their software build their own hardware;’ we need a lot of seriously good tooling to make that happen and I hope to build some of it.”
Learn more about Professor Nigam on his website or group page.
