The first version cost about fifty dollars. It was an RTL-SDR, the sort of USB radio hobbyists use to explore the airwaves, paired with a Raspberry Pi. The hardware was cheap enough to scatter around. The software was early. The business model was missing. Yet for Alex Wulff and his two Harvard friends, the prototype answered the important question: a network of modest radios, coordinated by good code, could begin to perform work normally assigned to scarce and expensive equipment.
The idea became Distributed Spectrum, the New York company Wulff runs with co-founders Isaac Struhl and Ben Harpe. Its sensors listen for radio activity, process data close to where it is collected, identify signals, and turn that analysis into an alert or a map. The use cases include nearby drones, cellular devices, jammers, and unfamiliar waveforms. The premise is simple enough to say in one breath: put capable software on inexpensive hardware, then deploy enough of it to see a radio environment that one box cannot.
Wulff has been circling this problem for years. He grew up near Syracuse, built electronics, entered science fairs, and arrived at Harvard to study electrical engineering. Before he was a defense-tech CEO, he was the maker who wanted other people to understand why radios worked. In 2019, while still a student, he published Beginning Radio Communications: Radio Projects and Theory, a 223-page guide that moves from antennas and modulation to weather-satellite images and handheld transceivers. The method was hands-on and qualitative. Build something, watch what happens, and let intuition catch up with theory.
The fifty-dollar question
That instinct followed Wulff into industry. During a 2019 Lockheed Martin internship in Syracuse, he worked on ground-based and airborne radar projects, including the patterns of energy a radar array emits as it scans. He arrived thinking of himself as hardware-oriented. He left, he said, with “a new respect for signals and simulation.” He also learned to ask engineers questions he had initially kept in his head. Online research could provide definitions; conversations supplied the intuition.
The experience exposed a productive tension. Large defense programs could coordinate work at a scale that demanded structure, process, and specialized teams. They also moved differently from a group of students with consumer components on a desk. Wulff later worked on sensor systems and electronic-warfare technology at Raytheon and Lockheed Martin. The technical domain stuck. So did his desire to work faster.
A gap year pointed toward the field
Wulff met Struhl near the beginning of freshman year through frisbee and early computer-science projects. Struhl's roommate Harpe completed the trio. By the time the pandemic pushed Harvard classes online, all three had discussed starting a company. Rather than spend their senior year on screens, they took time off in 2020 to build together. Wulff brought a proposed thesis question: could cheap commercial radio monitors, working as a network, approach the capabilities of far more expensive sensing systems?
They hacked for a few months and reached a convincing yes. Then came the wonderfully awkward part of applied science: gathering data. They built a few dozen sensors and asked classmates to keep them in dorm rooms. Wulff and his co-founders walked around campus talking into handheld radios so their algorithms could learn what those transmissions looked like across the network. “We had some very funny looks from some of our classmates,” Wulff later recalled. Another early session involved walking around town and transmitting for four or five hours at a stretch.
At first, they imagined commercial wireless customers. About six months into the company, an online search led them to Mad Hacks: Fury Code, a National Security Innovation Network competition. Conversations there changed the market. Operators described vehicles losing communications with no easy way to tell whether a radio had failed or an adversary was jamming it. Distributed Spectrum entered with its rough, low-cost prototype and won the $25,000 grand prize in February 2021 for detecting radio-frequency attacks on vehicles in real time.
“Our more general thesis as a company is to develop this modular, hardware-agnostic software platform.”Alex Wulff, while at Harvard
The customer is part of the system
Winning a hackathon made the need visible. It did not make the government easy to sell to. The founders learned that the person with the urgent operational problem might have no purchasing authority. Program offices, contracting shops, funding pathways, and end users each occupied a different point in the circuit. A promising federal grant application made in October 2020 was not awarded until January 2022. The waiting was a practical course in how public procurement keeps time.
Wulff has been unusually candid about that education. In 2023, he described the “valley of death” less as an abstract funding gap and more as the work of finding the real decision-makers and earning their trust. Early SBIR and STTR awards paid for development, but the company still had to learn how a small research project could become an operational purchase. Technical performance mattered. So did positioning the product so a near-term contract would not narrow the long-term platform.
The company's answer was mission-driven business development. The founders would bring a prototype to users, listen to what a mission required, change the product overnight, and return with something better the next day. That loop let seven people close more than $7 million in contracts during a 60-day stretch. By March 2025, Distributed Spectrum said it was on contract with the Army, Air Force, Navy, Special Operations Command, and the intelligence community.
days to close more than $7 million in contracts.
The company credited a seven-person team, founder-led sales, and an iterative demo cycle built around specific missions.
Radio has a distribution problem
The economic argument underneath the product is about coverage. Traditional radio-frequency sensing systems can cost more than a million dollars and require trained specialists nearby. A small Distributed Spectrum field unit was described in 2025 as containing roughly $1,500 to $2,000 in commercial hardware, including a software-defined radio and an Nvidia Jetson computer. The point is not that one inexpensive device matches every capability of a specialized platform. It is that many devices can occupy places where a million-dollar box never will.
When lower cost changes the map
Illustrative scale based on publicly described component costs. Bars use a compressed comparison for readability.
Distribution creates its own technical constraints. A sensor on a backpack, drone, buoy, vehicle, or remote platform cannot depend on abundant bandwidth or a rack of servers. It must process locally, use limited power, and deliver an answer without asking an expert to interpret a waterfall of raw waveforms. Distributed Spectrum combines machine learning with classical signal processing because neither generic AI tools nor rigid signal libraries solve the whole problem. The models must run fast, fit on constrained computers, and adjust as waveforms change.
That architecture found a vivid expression in the first AUKUS Electronic Warfare Challenge. Distributed Spectrum won the $150,000 U.S. prize with a maritime concept: scatter low-cost sensors across buoys and land platforms to form what Wulff called “one big radio sensing mesh.” Instead of searching the Pacific with a handful of scarce assets, the network could flag activity across a broader area and help distinguish one suspicious vessel from ordinary traffic.
The company has also said a partner in Ukraine strapped some of its sensors to drones to locate jammers and concealed troops. Other product paths include a sensor carried in an operator's backpack, stationary coverage around a base, and alerts for nearby drone or cellular activity. Each configuration expresses the same principle. Put the intelligence close to the antenna, then send the decision downstream.
“Radio signals are everywhere so we need sensors everywhere.”Distributed Spectrum's Series A announcement
Scaling the signal, keeping the loop
In March 2025, Distributed Spectrum announced a $25 million Series A led by Conviction, Shield Capital, and Nat Friedman, with participation from Felicis, XFund, and technology and defense operators. Retired Army General Stanley McChrystal invested and became an adviser. Conviction founder Sarah Guo backed the round after the company had moved beyond research awards into contracts and deployments. The network around Wulff now spans engineers, investors, military users, program offices, and founders who have learned other hard markets.
The capital arrived with a scaling problem. A few months before the round, the company had seven engineers. By the announcement, it had doubled to 14 people, added product and operations roles, and moved into a larger New York office. The founders said they wanted to deliver existing contracts, deepen research in embedded machine-learning signal processing, and make their path to market more repeatable.
The cultural language remains that of a lab with deadlines. The founding team has written about “extreme ownership,” working together in person, and hiring engineers who can become fluent across embedded systems, statistics, signal processing, and machine learning. Wulff's path offers a useful model for that kind of range. He moved from circuits to radar simulation, from a book for beginners to government acquisition, and from a thesis experiment to executive work without treating the technical vocabulary as someone else's department.
First invisible, then actionable
There is a quiet continuity to the career. A radio textbook helps a beginner notice a world already passing through the room. A networked sensor helps an operator notice the signal that matters inside that world. Both projects turn something abstract into an object for judgment. Wulff's work has changed in stakes and scale, but not in its basic fascination.
The next test is whether Distributed Spectrum can preserve its feedback loop while delivering more hardware, supporting more missions, and navigating larger programs. Cheap components alone are easy to copy. Durable value will live in the models, the deployment data, the integrations, and the trust accumulated with users. Wulff's stated ambition is to make this kind of sensing ubiquitous across services and wide areas. In his blunt formulation, “The only solution is to automate some of this.”
That ambition began with three friends walking around Harvard, radios in hand, while small boxes listened from dorm rooms. The experiment looked strange because the map it promised did not yet exist. Now the map is the company: many inexpensive points, learning together, trying to make the invisible legible before someone needs to ask what is out there.