I’m a medical doctor and founder of a small startup called Craniolabs. A month ago, my cofounders and I started working on a painfully obvious problem in medical AI: the lack of diverse, high-quality, ethically sourced medical imaging data, especially from underrepresented regions like Africa, Asia, and the Middle East.

Everyone in the field talks about bias in datasets, but very few seem to be solving it at the root. So we tried.

We signed memorandums of understanding with hospitals in Egypt and Dubai. We got access to radiology departments and de-identified DICOM archives. We partnered with radiologists. We integrated NVIDIA MONAI to streamline annotation. And then… we reached out to over 30 diagnostic AI startups who we thought would jump at the chance to access better data.

Almost no one replied.

Some opened our emails 3–4 times. A few asked for more info. One or two made it to pricing discussions. But the reality is: most teams either weren’t ready to buy, didn’t have budget, or were hesitant to engage in anything that looked operational.

Here’s what we learned: • Most diagnostic AI startups are resource-strapped, even if well-funded • Everyone wants clean, diverse data, but no one wants to manage the plumbing • Many are still stuck using NIH ChestXray or CheXpert and don’t trust third-party data easily • Even at small scale, hospitals need very clear legal, ethical, and financial frameworks to move data

We’re now restructuring toward a subscription-based model with ready-to-use curated batches, compliance built in, and optional annotations. But the lesson stuck: getting the first few customers is way harder than building the product.

Curious if others here have faced something similar, whether in healthcare, infra, or AI. If you’re building in this space or just have thoughts, I’d love to hear how you’d approach this.

(More about what we’re doing at: https://craniolabs.tech)