Wedge Is Building the Air Traffic Control Layer for Healthcare AI

The Macro: Healthcare AI Is Everywhere, and Nobody Is Driving

I want to paint a picture of what healthcare AI adoption actually looks like right now, because the press coverage and the reality are very different things.

The press coverage says healthcare is being transformed by AI. Radiology, pathology, clinical decision support, patient intake, claims processing, prior authorization. Pick a department and there is an AI startup pitching it. The market for AI in healthcare was around $20.9 billion in 2024 and is on track to hit something like $148 billion by 2029, depending on which research firm you trust. Those are real numbers reflecting real deployments.

The reality on the ground is messier. A large health system might have a dozen AI tools deployed across different departments, each purchased by a different stakeholder, each running on different infrastructure, each with its own compliance story. The radiology AI was approved by the CMIO. The billing AI was brought in by revenue cycle. The patient messaging bot was a project from digital health. Nobody has a unified view of what is running, what data it is touching, or whether any of it is performing as promised.

This is the classic enterprise middleware opportunity, and it shows up in every industry wave. Salesforce created it for CRM. Palantir created it for government data. ServiceNow created it for IT operations. When a category gets hot enough that every department starts buying its own tools, someone has to build the control layer.

Healthcare AI is at exactly that inflection point. The tools exist. The budgets exist. The governance layer does not. Companies like Aidoc and Viz.ai are building excellent point solutions for specific clinical workflows, but nobody is building the platform that sits above all of them. That gap is what Wedge is going after.

The Micro: Stanford AI Meets NVIDIA Engineering, Two Founders, One Very Specific Bet

Wedge was founded by Devraj Gopal and Steven Segawa. Gopal is an ex-Stanford AI researcher who worked on health AI at both Stanford and Johns Hopkins Medical. Segawa previously served as Lead ML Data Scientist at NVIDIA, with stints at Rivian, SoFi, and Intel. They came through Y Combinator’s Summer 2025 batch and are based in San Francisco.

Two people. That is the entire team right now. And I think that is fine for this stage, because the product they are building is fundamentally an infrastructure play, not a feature play. The question is not “can they build it” but “can they get health systems to adopt it before someone bigger moves in.”

The product positions itself as the operating system for healthcare AI agents. Deploy agents, set governance rules, monitor performance, scale across the organization. Think of it less as another AI tool and more as the layer that manages all the other AI tools. The Palantir comparison in their own positioning is apt. Palantir did not build the data. Palantir built the system that made the data usable across an organization. Wedge is trying to do the same thing for AI agents in healthcare.

What makes this timing interesting is HIPAA. Healthcare is not like fintech or e-commerce where you can move fast and apologize later. Every AI agent touching patient data needs audit trails, access controls, and compliance documentation. Most health systems are stitching this together manually. That is not sustainable at the current pace of AI adoption, and it creates a genuine wedge (pun fully intended) for a purpose-built governance platform.

The competitive landscape is thin in this specific niche. There are plenty of healthcare AI companies and plenty of AI governance companies, but the intersection of the two is mostly empty. Datadog monitors infrastructure. Arize monitors ML models. Neither is purpose-built for the specific regulatory and operational requirements of healthcare AI agents. Wedge is betting that healthcare is different enough to warrant its own platform, and based on what I have seen in the space, I agree.

The founding team combination is strong. You want someone who understands the clinical AI landscape deeply (Gopal) and someone who has built production ML systems at massive scale (Segawa at NVIDIA). Those two skill sets together cover the technical surface area this product needs.

The Verdict

I think Wedge identified a problem that is going to get dramatically worse before it gets better. Healthcare AI adoption is accelerating, governance infrastructure is lagging behind, and the regulatory pressure to close that gap is only increasing.

What I would watch at 30 days: are health systems actually willing to adopt a third-party governance layer, or do they want to build this capability internally? The “build versus buy” conversation in healthcare IT is notoriously slow and politically charged.

At 60 days: integration depth. The value of this platform scales directly with how many AI tools it can actually govern. If the integration list is short, the value proposition narrows quickly.

At 90 days: whether the Palantir positioning helps or hurts. In some healthcare circles, Palantir carries baggage. The comparison communicates ambition, but it might also communicate a sales process that takes 18 months and a government contract.

Two founders, a clear gap in the market, and the right technical backgrounds. I am paying attention.