INVESTMENT THESIS

Where Software Meets the Physical World

I invest in technologies that sit at the boundary between software and the physical world – where progress is constrained by physics, systems, and deployment, not just code.

Over the past decade, software ate software. The next phase of technological progress is about software eating the physical world: machines, infrastructure, manufacturing, logistics, and the systems that move atoms, energy, and materials through reality.

This shift favors a different kind of company:

• Hard to build
• Hard to copy
• Grounded in real-world constraints
• Improved through deployment, not demos
• Often unsexy, but enormously valuable

My focus is on embodied intelligence, automation, and industrial systems – technologies that turn intelligence into action and embed learning directly into the physical world.

Category I: Embodied Intelligence & Automation

Learning Systems in the Real World

We are entering an era where intelligence is no longer confined to screens or decision-support tools. It is being embedded directly into machines that perceive, decide, and act in the real world.

The convergence of AI, robotics, automation, and sensing is enabling a new class of systems that operate in dynamic, unstructured environments – warehouses, factories, construction sites, logistics networks, and field operations. These are environments where edge cases dominate, conditions change constantly, and success depends on tight feedback loops between perception and action.

This is not about humanoids for demos or autonomy for its own sake. It is about making physical work more productive, safer, and more scalable.

Why Now: From Software Intelligence to Physical Execution

Recent advances in machine learning made intelligence cheap and ubiquitous in the digital world. The next challenge is execution in the physical one.

Several forces are converging:

• Sensors are cheaper, better, and more ubiquitous
• Models are efficient enough to run at the edge
• Data is increasingly generated through real-world operation
• Interfaces between humans and machines are improving rapidly

Together, these shifts enable systems that learn from deployment and improve over time. The result is embodied intelligence that compounds – not because of clever algorithms alone, but because it is embedded in real operations.

What I Look For

In embodied intelligence and automation, I focus on companies that:

• Close the loop between sensing, decision-making, and action
• Learn through deployment in the real world
• Replace brittle, rule-based systems with adaptive ones
• Deliver measurable gains in productivity, safety, or cost
• Are constrained by engineering and operations, not just software talent

These companies tend to be slower to start, but far more defensible once they work.

Category II: Industrial & Physical Systems Infrastructure

Software-Defined Industry

As intelligence moves into the physical world, infrastructure becomes the bottleneck – and the opportunity.

Industrial systems were historically static, over-engineered, and slow to change. Today, they are becoming software-defined: instrumented with sensors, governed by data, and optimized continuously through learning systems.

This category includes the technologies that make embodied intelligence possible at scale:

• Industrial software and tooling
• Sensing and monitoring systems
• Materials and manufacturing technologies
• Energy- and compute-aware systems
• Infrastructure that must operate reliably in the real world

These are not consumer products. They are systems that industry depends on.

Why Now: Reality Is Becoming Programmable

Three structural shifts are underway:

1. Intelligence Is Becoming a System Primitive
   AI is no longer an add-on. It is being embedded directly into how physical systems are designed, operated, and optimized.
2. Deployment Is the Differentiator
   The hardest problems are not model accuracy, but integration, reliability, and scaling in messy environments. Companies that solve these gain durable advantages.
3. Constraints Create Moats
   Physics, supply chains, certification, and operational complexity slow competitors and reward teams that can execute end-to-end.

The best companies in this category look unglamorous early on. Over time, they become deeply embedded and extremely hard to replace.

What I Look For

In industrial and physical systems infrastructure, I focus on companies that:

• Improve how physical systems are built, operated, or maintained
• Replace manual, heuristic-driven processes with learning systems
• Are grounded in real customers and real deployments
• Deliver incremental value that compounds over years
• Sit at the intersection of software, hardware, and operations

These businesses often define categories quietly – and dominate them over time.

A Unifying Perspective

Across embodied intelligence and industrial systems, a common pattern emerges:

• Intelligence is moving closer to matter
• Learning happens through deployment, not abstraction
• Constraints are features, not bugs
• The most valuable systems are built slowly and deliberately

I am drawn to companies operating at the edge of what is possible – frontier engineering with commercial intent. Not science projects, not pure software plays, but technologies that work in the real world and get better because they are used.

This is where progress is hardest – and where it lasts.