Reflections from Our NatSec Hackathon
AI-powered lasers, auto-generated intelligence reports, and more...
This weekend Shield Capital hosted 500+ hackers, government leaders, students, engineers, founders, and investors at Silicon Valley's largest ever National Security Hackathon. The hackathon was sponsored by Shield Capital, along with the Stanford DEFCON Student Network (which was officially kicked off at the hackathon), Founders Fund, In-Q-Tel, Anduril, Stanford Gordian Knot Center, Primer AI, Radiant AI, and SkyFi.
Hackathon teams tackled problem sets curated by government stakeholders from SOCOM, United States Air Force, United States Navy, DoD Chief Digital and Artificial Intelligence Office, Defense Digital Service, Defense Innovation Unit (DIU), and Ukrainian warfighters focused on four main themes:
Digital drone resilience and operation
Next generation intelligence analysis
Computational logistics and reporting
AI-enhanced telemetry and sensing
In addition to curated problem sets, teams also had access to troves of data such as Russian and Chinese news and social media data, drone flight and collision data, and satellite imagery. Teams built everything from laser-powered counter UAS systems to automated intelligence report generation tools to levitating space crystals.
The weekend kicked off with talks from the CTO at Joint Special Operations Command and the CTO of the CIA and ended with remarks from Trae Stephens, co-founder of Anduril and partner at Founders Fund, and Raj Shah, Shield Capital’s managing partner and the former head of the Defense Innovation Unit. All the speakers had deep experience both in startups as well as in the government, and all of them emphasized the importance of fielding commercial technology in order to improve the US’s national security posture.
The event attracted hackers from many different backgrounds – military, startups, big tech, students. Some participants had no prior national security background while others had spent years steeped in the national security ecosystem.
Hackers and government mentors alike worked late into the night – the hacker Discord channel was active until at least 5AM and people stayed overnight in order to finish their projects in time!
I was struck by how powerful it was to have national security stakeholders from places like SOCOM, CDAO, SOCPAC, and DDS in the room as engineers worked on tangible solutions to the problems they proposed. Government mentors spent hours going back and forth with engineers providing feedback to help them build solutions that actually help solve problems the DoD is facing today. Typically, students and engineers at places like Meta and Google have pretty limited opportunities to interact with national security stakeholders, and vice versa. However, this hackathon showed the impact of combining the skills and knowledge from these different groups.
We were lucky to have some of the most technically savvy government representatives in attendance. Many of the government representatives in attendance previously worked at the Defense Digital Service (DDS, now part of CDAO), which is lovingly (if somewhat awkwardly) referred to as the “DoD’s SWAT Team of Nerds.” These folks are technologists trained in data and computer science. They’re called in to different DoD organizations to find ways to solve the DoD’s problems using data analytics and technology. Often the people in the DoD assigned to handle technological integration are not technologists at heart – sometimes they’re chosen simply because they have availability in their schedule. So, I felt very privileged to get to spend the weekend with some of the folks at the DoD who truly understand technology as well as the national security mission.
Some of the folks we had in attendance are directly engaged in the DoD’s new Replicator program, which is working to field small, attritable autonomous systems for air, ground, and sea operations. They were able to provide direct advice to participants about what is needed for autonomous system operation and protection. Another one of the government mentors at our hackathon is one of the technical leads for Task Force Lima, which works to integrate generative AI into the DoD. He was able to guide hackers through the realities of deploying generative AI in military environments. We even had a few Ukrainians on the front line available virtually to discuss topics like terminal guidance for autonomous systems and counter drone tactics with participants. Hackers got to learn from government experts who have truly lived the problems they were working to tackle and who really understand what it takes to build and deploy technology for national security use cases.
Overall, I was tremendously impressed with what can be built in 24 hours when the right people find themselves in a room together. Teams with little to no prior national security experience were able to shoot drones out of the sky, generate coherent intelligence reports based on multimodal data, trick AI deep fakes into revealing themselves, allow drones to communicate via lasers, and much much more.
Several of the government folks in attendance told me this hackathon was one of the most productive and impactful 24 hours of their recent careers.
The winning teams were:
First place: AI Laser Sentinel (Margarita Geleta, Michael LaFramboise, John Marmaduke, Justin Zelaya) built an AI-powered laser system that can shoot enemy drones out of the sky. You can follow their journey building this system here.
Second place: Overwatch (Brian Wu, Justin Lin, Ian Kim, Pavan Agareal) deployed computer vision models on drones.
Third place: Panopsis (Artur Kiulian, Artem Kalyta, Martin Bach, Philippe Brule) built a system to automatically generate SPOTREP reports based on satellite data, social media data, and audio data. and audio data.
I’ll also call out a few other teams I met that I was particularly impressed with (unfortunately I did not get the chance to meet everybody…if only there were more hours in a day!):
A team of UCSD and CMU PhD students developed a system that “poisons” an AI audio generation model in order to trick an AI audio deepfake scammer into revealing that it is an AI bot. Maybe one day a system like this can be deployed into all of our cellphones to prevent AI call scams.
A team of engineers developed a system that can analyze drone flight data in order to identify drones with similar flight patterns to track potentially nefarious activities (ex: a drone that drops off drugs to the same location each week)
A team of engineers and founders who built a system that enables drones to communicate with lasers to prevent them from being vulnerable to RF jamming.
Once again, I want to say thank you so much to the government stakeholders and sponsors who helped make this event happen. It was a hectic and exhilarating weekend, and I learned so much about what can be done for national security when enough smart and knowledgeable people are brought together.
For everyone who participated in the hackathon who wants to stay involved in the national security community, I highly encourage you to work at a startup building in national security (see, for example, Shield Capital’s portfolio companies), start your own company (and reach out to me if you do!), or find a way to join government service. Our country needs more talent working on solving the hardest and most important problems like those we tackled this weekend.
As always, if this hackathon inspired you or anybody you know to start a company building at the intersection of national security and commercial markets, please do not hesitate to reach out.