đ Engineers â want to meet these four founders and more? Apply to attend our September 16th SF Startup Showcase.
This group of founders is building for a world where software moves beyond the screen. Mariana is rebuilding mining and refining around software, EON is moving data transfer infrastructure to space, Generalist is building intelligence for robots, and Unlimited is rethinking how we design and construct the infrastructure itself. Its a theme weâve seen before; some of the biggest opportunities in technology now come from applying the speed, iteration, and intelligence of software to the physical world.
Charlie Horowitz, Founder of EON
Roughly 95% of the worldâs intercontinental data still travels through long fiber-optic cables on the ocean floor. These cables take years to deploy, cost hundreds of millions of dollars, and can be damaged by ship anchors, natural disasters, and deliberate sabotage. EON is building an alternative: skip the ocean and move the data through space.
Charlie came to the problem after starting his career at Palantir and later joining satellite manufacturer Apex as its tenth employee. At Apex, he kept hearing the same request from potential customers: they needed enormous amounts of data capacity without being completely dependent on subsea fiber. The concern was primarily geopolitical; in places like Taiwan, contingency planning assumes subsea cables would be the first infrastructure targeted in a conflict. After connecting with a major cloud infrastructure company over pricing on existing routes, Charlie realized they could solve for more than just redundancy. They could also move data dramatically cheaper.
The technology is, in Charlieâs most simplistic description, âspace lasers.â EON uses telescope-like optical ground stations to beam data to satellites with infrared light and back down again, targeting 2.4 terabits per second per link. The company just raised a $10.75M seed led by General Catalyst and a16z. When we spoke with Charlie, EON was just three months into operating formally, had seven employees, and is preparing to open roughly 15 roles across mechanical, electrical, embedded, controls, optics, and other disciplines.
If you dreamed of building âspace lasersâ as a kid, hereâs your chance. Come meet Charlie at the showcase.
Pete Florence, Founder of Generalist
Robots are remarkably good at doing the exact thing they were designed and programmed to do, but theyâre terrible at encountering a new object, environment, or task and figuring out what to do next. LLMs made the leap from narrow systems to general-purpose intelligence by scaling models, compute, and enormous datasets. Generalist is betting that the same transition is finally possible for robots.
Pete describes solving intelligence in the physical world as the âultimate act of engineering.â In Peteâs words, everything theyâre doing is âstrictly harder than making language models.â The challenges span across model training, data infrastructure, firmware, controls, mechanical engineering, and robotics simultaneously.
Generalistâs latest model, GEN-1, is trained largely from scratch using more than half a million hours of physical-interaction data. The team behind it has worked across OpenAI, Google DeepMind, and Boston Dynamics, and the company recently announced $400M in funding from Radical Ventures, 8VC, and USV among others. But one of Peteâs favorite company metrics has nothing to do with fundraising: Generalist has grown to 43 people, yet not a single technical employee has left in two and a half years. He attributes that in part to a culture where engineers develop deep expertise without being pigeonholed into one narrow problem forever.
If you want to work on the full stack required to give machines general intelligence in the physical world, come meet Pete at the showcase.
Juan Lozano, Founder of Mariana Minerals
AI infrastructure, electric vehicles, modern energy, and defense all depend on critical minerals like copper and lithium. The problem is actually getting them out of the ground. Mining and refining remain notoriously slow industries, with major projects often taking years to permit, build, and bring online. Mariana is trying to rebuild the minerals industry as a software-first operation.
Juan Lozano came to Mariana after building and scaling machine learning systems at Affirm and Kite. At Mariana, software isnât just an analytics layer sitting on top of a traditional mining company. Its internal platform, MarianaOS, spans the entire operation: software for building projects, running mines, and optimizing refineries. MineOS uses physics-informed world models and reinforcement learning to coordinate mining fleets, while PlantOS uses sensing and RL to continuously optimize mineral processing.
And theyâre already putting it to work in the real world. Mariana is operating Copper One in Utah and building Lithium One in Texas, with MarianaOS already deployed across both sites. The company says it compressed Copper Oneâs typically multi-year acquisition, permitting, and restart process into months, while Lithium One is on track to become its first designed-for-autonomy greenfield refinery less than three years after Mariana was founded. This month, Mariana announced a $310M Series B led by Khosla Ventures, bringing total parent and project capital raised to roughly $400M.
If you want to build software and AI where the output is measured in tons of copper and lithium rather than tokens, come meet Juan at the showcase.
Alex Modon, Founder of Unlimited Industries
Unlimitedâs manifesto opens with the line âWe used to build.â Their argument is that construction hasnât just failed to keep pace with softwareâit has developed incentives that actively resist speed. Cost-plus contracts, billable hours, rigid design freezes, and endless change orders reward caution and delay rather than better outcomes. Unlimited is rebuilding that system, with the goal of making physical infrastructure move at the pace of software.
Unlimited is an AI-native, vertically integrated engineering and construction company using small, engineering-led teams and software to actually deliver factories, data centers, energy infrastructure, and other large projects. Unlimited uses AI to generate hundreds of thousands of design permutations, drastically cutting feedback loops, iteration time, and overall cost.
The approach is already producing real results. Unlimited has compressed pre-construction engineering that traditionally took six months into a few weeks and identified redesigns that cut projected capital costs by more than 50%. The company raised a $12M seed co-led by Andreessen Horowitz and CIV. When we had dinner at Unlimitedâs office this summer, the pace was tangible: team members arrived from calls, disappeared to run models, and came back to the tableâwelcoming, but visibly eager to get back to building.
If you want to help restore our ability to build ambitious things in the physical world, come meet Alex at the showcase.
Apply to attend our next SF Startup Showcase and meet these founders in person.