Josh Miller, CEO and co-founder of Gradient Health, joins me to share how a Durham-area startup became the go-to source of medical data for AI developers around the world.

Josh traces the path from his first company, FarmShots, a satellite imaging startup acquired by Syngenta, to Gradient, which began as an attempt to build cancer-detection AI and turned into something bigger once the team saw how hard it was to get the data. The conversation covers why de-identifying medical data is so much harder than removing a name, how foundation models changed the value of labeled data, and why expert-curated datasets are becoming the next frontier.

Gradient's numbers are remarkable: roughly 40 million dollars in revenue this year with about 40 people, on only about 9 million dollars raised. Josh credits a culture that is willing to talk openly about what isn't working, along with AI-native engineering that has made his best engineers dramatically more productive. He also talks about go-to-market when customers are racing a 20 million dollar training run, his company's moat, and a personal reminder for founders to take care of themselves on the long road.

Timestamps
00:00 Welcome to NC Tweener Talks
00:25 Thanks to our sponsors
01:35 Scot's intro: portfolio wins you can't talk about
02:30 Why Tweener invested in Gradient Health
04:55 Welcome Josh Miller
05:55 FarmShots: satellite imagery for farms
07:45 A lever through time
08:30 The Syngenta exit and Duke roots
09:25 Never again… then startup #2
10:40 Founding Gradient before ChatGPT
11:30 From building cancer AI to supplying the data
13:30 De-identification and the 87% problem
14:30 Paying hospitals back
15:20 Weak labels vs. strong labels
16:55 First contract and the revenue curve
19:20 Fundraising: $9M raised in total
20:35 Getting pulled by the market
21:55 Atlas 2 and multimodal data
23:00 Picks and shovels vs. building the model
23:55 40 people, $1M per head
25:30 Post-training and expert-curated data
27:35 Culture: talking about problems
28:55 AI-native engineering and the 100x engineer
30:15 Go-to-market in the AI data space
32:10 Gradient's competitive moat
33:10 Founder lesson: take care of yourself
34:15 Wrap and credits
---
Where to Find Josh Miller:
LinkedIn: https://www.linkedin.com/in/joshmichaelmiller/
Gradient Health: https://gradienthealth.io/

Where to Find Scot Wingo:
LinkedIn: https://www.linkedin.com/in/thescotwingo/
Tweener Times: https://www.tweenertimes.com/
X: https://x.com/scotwingo

---

This episode of Triangle Tweener Talks is hosted by Scot Wingo, presented and produced by NC Tweener Fund, with creative assets and design support from Walk West.

We couldn’t share posts like this without our amazing sponsors:

Platinum:
NC IDEA: https://ncidea.org

Gold Sponsors:
- Balentine: https://www.balentine.com/triangle-entrepreneurs
- EisnerAmpner: https://www.eisneramper.com
- Robinson Bradshaw: https://www.robinsonbradshaw.com

Silver Sponsors:
- Automated Consulting Group: https://automated.co
- Bank of America: https://business.bofa.com/en-us/content/technology-industry-group.html

------
Triangle Tweener Talks is sponsored by:

• Atomic Object: https://atomicobject.com/

Craig Brown, a field engineer at Quantexa and the solo builder behind Blind Oracle, joins us as Redacted's first inbound guest.

Blind Oracle is a trust and audit layer for AI agents doing business with other agents. It issues agent passports and records delegated work to public ledgers, so anyone can verify what an agent did without having to trust the people who run it. Craig screen-shares the software factory he built to develop it over the past two years. Signals from email, voice notes, social media, and research papers feed planning models, and a governance layer scores the system on alignment, autonomy, durability, and more.

We dig into the questions every serious builder eventually hits. How do you evaluate a system once you've stopped reading its code? How do you build real trust in agents when the underlying models keep changing? And what should true north be when the product is early, and revenue hasn't arrived yet?

Craig is candid about all of it, including his bet that agent-to-agent commerce is coming and that the agents themselves will decide who they trust.

Timestamps:
00:00 Cold open: selling to the robots
00:44 Welcome to Redacted
01:00 Our first inbound guest
02:40 Craig's background: Quantexa and decision intelligence
04:05 From Lockheed Martin to Mantas
05:30 Solution engineering vs. forward deployed engineering
06:15 Building on the side: "the baby" turns one
07:40 Three ways to demo
11:10 What Blind Oracle does: agent counterparty risk
12:15 Agent passports and inter-agent commerce
12:55 Proof on-chain: Nostr and Base
14:40 Inside the software factory: signals, plans, builds
17:05 Feeding an article into the harness
23:35 How do you eval the factory?
24:15 Multi-agent deliberation
25:50 The governance scorecard
29:00 The honest failure: no revenue yet
29:50 Go-to-market for agents
31:00 How "decision confidence" gets measured
33:40 Building trust in an agent system
37:20 True north: the factory or the product?
39:50 The bet on agent-to-agent commerce
42:35 Research papers as signals
44:15 Model benchmarks and routing
46:30 Where to find Craig

---

Show notes from the episode: https://github.com/instanttaylor/redacted-podcast

Where to Find David:
LinkedIn: https://www.linkedin.com/in/davidshaner/

Where to Find Taylor:
LinkedIn: https://www.linkedin.com/in/taylorcotner/

More about Offline: https://www.linkedin.com/company/offline-media-inc-/

---

This episode of Redacted is hosted by David Shaner and Taylor Cotner, and presented and produced by NC Tweener Fund.

We couldn’t share posts like this without our amazing sponsors:

Platinum:
NC IDEA: https://ncidea.org

Gold Sponsors:
- Balentine: https://www.balentine.com/triangle-entrepreneurs
- EisnerAmpner: https://www.eisneramper.com
- Robinson Bradshaw: https://www.robinsonbradshaw.com

Silver Sponsors:
- Automated Consulting Group: https://automated.co
- Bank of America: https://business.bofa.com/en-us/content/technology-industry-group.html

------
Triangle Tweener Talks is sponsored by:

• Atomic Object: https://atomicobject.com/

Michael Boyarsky, CEO and co-founder of Extellis, joins me to talk about turning more than a decade of Duke metamaterials research into a satellite imaging company.

Michael explains what metamaterials are and why heat, not cold, is the real challenge for satellites in orbit. He also describes the conference moment when a competitor bragged about 2 percent uptime, and he thought the industry could do far better. He walks through the path from a DARPA-funded ground prototype to a spreadsheet that made the business case clear, and from there to spinning out of Duke with help from Duke New Ventures.

The conversation also covers how Extellis is getting to orbit without building its own satellites. The company is partnering with Astro Digital and riding on the ecosystem SpaceX made possible. Michael shares the details of its oversubscribed 6.8 million dollar seed round led by Oval Park Capital and the upcoming Pathfinder satellite. He closes with advice for researchers who are thinking about starting a company: begin with a real market need, not a technology looking for a home.

This is the Tweener Fund's first space investment, and it's a great listen for anyone curious about deep tech, university spinouts, or the new space economy.

Timestamps
00:00 Welcome to NC Tweener Talks
00:25 Thanks to our sponsors
01:35 Scot's intro: the Tweener Fund's first space investment
03:30 Welcome Michael Boyarsky
04:00 From Boston College physics to Duke's metamaterials lab
06:35 What is a metamaterial?
08:05 Negative refraction and the invisibility cloak
10:45 From lab demos to satellite imaging
11:45 Why heat is a problem in space
13:45 The 2% uptime moment
15:35 DARPA prototype: could we hit 20-25%?
17:00 How you build a metamaterial
20:10 "The company started in a spreadsheet"
21:55 Duke New Ventures and learning to run a company
24:00 Customer discovery in deep tech
26:45 Target markets beyond defense
29:30 Why Extellis needs its own satellite, but not its own factory
32:25 SpaceX and the space ecosystem
36:40 Fundraising: Oval Park Capital and Shell Ventures
38:55 The pathfinder: Astro Digital and a ride on a Falcon
41:05 Pathfinder goals, size, and orbit
46:00 Advice for deep tech founders and wrap
-----

Where to Find Michael Boyarsky:
LinkedIn: https://www.linkedin.com/in/michaelboyarsky/
Extellis: https://www.extellis.com/

Where to Find Scot Wingo:
LinkedIn: https://www.linkedin.com/in/thescotwingo/
Tweener Times: https://www.tweenertimes.com/
X: https://x.com/scotwingo

---

This episode of Triangle Tweener Talks is hosted by Scot Wingo, presented and produced by NC Tweener Fund, with creative assets and design support from Walk West.

We couldn’t share posts like this without our amazing sponsors:

Platinum:
NC IDEA: https://ncidea.org

Gold Sponsors:
- Balentine: https://www.balentine.com/triangle-entrepreneurs
- EisnerAmpner: https://www.eisneramper.com
- Robinson Bradshaw: https://www.robinsonbradshaw.com

Silver Sponsors:
- Automated Consulting Group: https://automated.co
- Bank of America: https://business.bofa.com/en-us/content/technology-industry-group.html

------
Triangle Tweener Talks is sponsored by:

• Atomic Object: https://atomicobject.com/