ARMONEX: Neuro‑symbolic clinical intelligence for ambulatory oncology
Clinical AI framework designed to anticipate deterioration in oncology patients at home by combining medical‑grade biosensors, predictive models, and explainable clinical knowledge graphs. Its goal is to turn continuous vital sign streams into clinically meaningful, contextual alerts that clinicians can trust, audit, and integrate into everyday decision‑making.
Tech stack:
What ARMONEX does
ARMONEX introduces a neuro‑symbolic architecture that combines deep learning with structured clinical reasoning to deliver three core capabilities eliminating the «black box» effect:
Capture
Capture continuous signals from medical‑grade patches and wearables (heart rate, respiratory rate, temperature, blood pressure, oxygen saturation, and others) in the context of the patient’s active treatment.
Analysis
Analyze in real time vital signs and its deterioration. Automatically filter noise and artifacts, reducing alarm fatigue and prioritizing only those patterns that represent meaningful clinical risk.
Interpret
Generate traceable, medically grounded explanations for each alert, based on clinical ontologies and terminology standards, making AI outputs understandable and auditable by clinicians.
Neuro-symbolic artificial intelligence
“ As the European Data Protection Supervisor notes, neuro‑symbolic AI can “merge unstructured medical data with structured knowledge bases like medical guidelines and ontologies,” enabling pattern recognition and logical, medically grounded reasoning in a single system. This is precisely the kind of auditable AI architecture healthcare now needs.”
EDPS
My Story
My story sits at the intersection of advanced AI research and real‑world healthcare. After two decades architecting large‑scale data and cloud ecosystems for global brands, I chose to focus my PhD on one question: how can we make AI in oncology both predictive and clinically trustworthy.
Today I lead the design of ARMONEX, a neuro‑symbolic early warning framework that combines continuous biosensors, deep learning, and clinical knowledge graphs to anticipate deterioration in breast cancer patients at home. My work is about turning cutting‑edge models into transparent, regulatory‑grade tools that oncologists can understand, audit, and eventually rely on in daily practice.
