Physicians Are Now Building Their Own Healthcare AI Systems
Sofya, a Miami-based startup, is betting that doctors will increasingly want to build custom AI workflows themselves rather than waiting for vendors to deliver them.
The shift is already happening. Physicians are using AI coding tools between patient visits to prototype healthcare workflows tailored to their own practices. Igor Couto, Sofya's founder and CEO, calls this "vibe coding" - and says it's becoming surprisingly common.
"We are seeing a movement when physicians and health systems are using AI to program by themselves," Couto said in a recent interview.
From Automation to Clinical Reasoning
Healthcare's first wave of AI adoption focused on administrative tasks. Hospitals deployed AI scribes for documentation, tools for scheduling and claims management, and systems to handle operational workflows. Physicians embraced these tools because they saved hours of paperwork.
But some doctors wanted to go deeper. They began asking whether AI could help with the harder problem: reasoning through complex medical decisions in real time.
That shift pushed Sofya toward clinical reasoning itself, particularly in specialties like oncology. The company recently completed a program through Mayo Clinic's platform, developing AI models for precision oncology focused on prostate cancer treatment decisions.
Sofya structures patient data and clinical guidelines into reasoning systems designed to help physicians navigate difficult treatment decisions. The company is now deploying across hospitals and clinics in Latin America while beginning early implementations in the United States.
The Problem With Physicians Building Alone
Couto noticed something telling while watching doctors attempt to build AI systems on their own: most projects ran into the same problems.
Security issues emerged. Interoperability failed. Long-term maintenance became impossible. Physicians had the clinical expertise but lacked the infrastructure to sustain custom AI workflows.
That realization reshaped Sofya's strategy. Instead of the company specializing workflows for individual hospitals, Sofya is building what it calls a programmable clinical studio - an environment where hospitals and physicians can customize AI workflows on top of Sofya's underlying infrastructure.
"Like Claude Code, but for physicians," Couto said, referring to the AI coding tool.
How It Works in Practice
The model gives healthcare systems flexibility without forcing them to build from scratch. A Miami hospital system could design workflows tailored to its patient population. A specialist could create reasoning models deployable across clinics worldwide.
Hospital Israelita Albert Einstein, one of Brazil's leading hospital systems, has been validating Sofya's clinical workflows. In the United States, the company is in discussions with the University of Miami and hospital systems in Texas.
The programmable studio approach reflects a broader shift in how AI is being adopted in healthcare. Rather than vendors dictating how AI should work, healthcare organizations are demanding tools that adapt to how they actually practice medicine.
For more on AI for Healthcare and how generative AI and LLM systems are being applied to clinical settings, see our training resources.
Your membership also unlocks:
