How Microsoft’s Julian Isla is Transforming Rare Disease Diagnosis with AI

When a child falls seriously ill, every second spent without answers feels like a lifetime. For Microsoft software developer Julian Isla, this reality struck hard when his infant son, Sergio, began suffering from unexplained seizures. After a harrowing year of misdiagnoses and confusion, the final diagnosis came: Dravet syndrome, a rare and severe neurological disorder. The experience revealed deep cracks in the healthcare system’s ability to recognize and diagnose rare conditions quickly.

Motivated by his

Foundation 29’s early milestone was creating a simple but effective diagnostic AI tool. Fast-forward to 2023, and they introduced DxGPT — a major leap forward. Built using advanced language models like GPT-4o and hosted on Microsoft Azure, DxGPT represents a new era of healthcare assistance.

DxGPT operates by analyzing symptoms described by patients or caregivers and suggesting possible diagnoses within minutes. It prioritizes privacy by not storing personal data or requiring user identifiers. While it doesn’t replace a physician’s expertise, it empowers families and doctors with quicker, more informed starting points.

Today, DxGPT is freely available online, providing hope and practical support to countless families navigating complex, life-altering health challenges.

What Undercode Say:

Julian Isla’s work through Foundation 29 reflects a growing, urgent trend: the intersection of AI technology and personalized healthcare. His project stands as a real-world example of how AI can drastically reduce the time between symptom onset and accurate diagnosis — a particularly crucial factor in rare diseases, where delays can mean irreversible damage or loss of life.

Analysis shows that around 1 in 17 people will be affected by a rare disease at some point in their lives. Yet, it often takes years for a diagnosis, during which patients undergo multiple tests, treatments, and emotional roller coasters. DxGPT targets this problem by offering preliminary diagnostic suggestions within minutes, potentially cutting down years of uncertainty into days or weeks.

The technology behind DxGPT — using large language models trained on a mixture of public and private medical datasets — also highlights the importance of curated, high-quality data in healthcare AI. Without accurate training data, even the smartest model risks dangerous inaccuracies. Here, Foundation 29’s collaboration with healthcare institutions to secure reliable data sources stands out as a strong practice in responsible AI development.

Privacy remains a huge concern in medical applications, and DxGPT seems well-positioned by not collecting or storing any user identifiers, maintaining GDPR compliance and patient trust. This model could serve as a benchmark for future medical AI tools looking to balance innovation with ethical considerations.

Moreover, DxGPT’s open accessibility is a significant move toward democratizing healthcare. While many diagnostic tools remain locked behind paywalls or require healthcare provider access, DxGPT allows anyone with an internet connection to get preliminary guidance, potentially saving lives, especially in underserved or remote regions.

Statistical support: According to a 2023 study published in Orphanet Journal of Rare Diseases, AI-based diagnostic tools have improved diagnostic rates by up to 25% compared to traditional symptom-based workflows alone. While not flawless, AI companions like DxGPT could radically shift how primary symptoms are triaged worldwide.

Foundation 29’s case also hints at a deeper cultural shift within tech giants like Microsoft: personal experience leading to impactful innovation. Isla’s personal tragedy did not just birth a technological solution; it opened the door for more empathetic, patient-centered AI developments within large corporations.

In short, DxGPT isn’t just another AI tool — it’s a glimpse into the future of healthcare: faster, smarter, and, most importantly, more human.

Fact Checker Results:

DxGPT was confirmed to use GPT-4o and o1 models, sourced from a blend of public and private healthcare datasets.
Foundation 29 was officially co-founded by Julian Isla in 2017, primarily driven by personal experience.
DxGPT’s policy of not storing personal data aligns with GDPR and modern healthcare privacy standards.