NEW JERSEY — A 34-year-old New Jersey man built an AI-driven software platform during his mother's 76-day battle with stage 4 duodenal adenocarcinoma that he said detected errors in her treatment plan and flagged a pulmonary embolism on Christmas Day. Pratik Desai, who had never served as a caregiver before his mother's illness, coded the system by combining multiple artificial intelligence tools to monitor her daily medical data and coordinate her ongoing cancer care.

Desai's mother was diagnosed with the advanced cancer, which affects the small intestine. After the diagnosis, doctors planned to discharge her without scheduling an oncology appointment, according to Desai. In response, he used AI to research every hospital in New Jersey for available oncology appointments. He also asked AI systems to help him develop a working understanding of stage 4 duodenal adenocarcinoma so he could engage more effectively with her medical team.

The platform Desai constructed relied on a daily workflow that began with exporting medical data from the Epic electronic health records system into NotebookLM, an AI tool he used to synthesize his mother's symptom histories and medical record data. He then fed the synthesized output into the AI model Claude to interpret the information and prepare questions and context for upcoming medical appointments. Together, these tools formed the core of the software platform he developed to track potential issues in her care plan.

On Christmas Day, Desai observed changes in his mother's walking, breathing, and speech. Medical staff did not return his calls that day. He input her symptoms into the AI tool, which identified complications consistent with a pulmonary embolism. The platform also detected what Desai described as errors in his mother's broader cancer care plan, though those findings have not been independently verified by medical professionals.

His mother spent 67 of her 76 days after diagnosis as an inpatient. "I was by her bedside from 5 a.m. to 10 p.m. every single day throughout the 76 days she was alive," Desai said.

"I believe the AI saved her life three times," Desai said. All accounts of the platform's detection capabilities originate from Desai, who served as both the software's sole developer and his mother's primary caregiver throughout her illness. No independent medical expert has publicly assessed the tool's findings or its accuracy in identifying the errors and conditions Desai described.