Our technology
Safe. Precise. Scalable.
The technology behind every triage.
Probabilistic reasoning
A Bayesian inference engine weighs thousands of conditions at once — not a fixed decision tree.
Certified medical device
MGP is CE-marked as Class IIb medical device software under Regulation (EU) 2017/745.
Clinically validated
Built on 160,000+ hours of physician expertise and validated in peer-reviewed studies.
How it works
The neuro-symbolic reasoning driving every interview
One engine combining language models and medical reasoning — ensuring every triage decision is explainable, auditable, and clinically grounded.
The interview
Inside the technology
From reading and mapping symptoms to delivering a next-step recommendation, each component plays a distinct, explainable role.
01 · Medical Knowledge Base
Grounding every assessment in verified medical evidence
Each input is fed through a structured map of 3000+ conditions, symptoms and risk factors. These are connected by over 1,000,000 relations and standardized against SNOMED CT and ICD-10 taxonomies.
)
02 · Natural Language Processing
From vague symptom descriptions to clinical evidence
The NLP layer reads free-text patient input and maps symptoms to structured clinical concepts. It’s available in 20+ languages.
)
03 · Bayesian Inference Engine
Adaptive interview logic assess the next best question
A Bayesian algorithm scores every condition against the evidence, adjusting the next question in real-time. An adaptive interview, never a fixed script.
)
04 · principled ai reasoning
Turning probability into a defensible case
Every candidate condition is synthesized from the language & reasoning models into one coherent, evidence-linked case. Reviewable like a clinical argument, never a bare percentage.
)
05 · AI guardrails
Ensuring safety through strict clinical protocol
Every recommendation is checked against clinical escalation rules, and every input is screened for prompt injection or misuse. Rules can only raise the triage level, never lower it.
)
06 · Explainable output
Delivering transparent and traceable recommendations
Every finding, probability and rule activation is logged in a path from evidence to output. Every recommendation is traceable so there’s never a black box effect.
)
Medical Guidance Platform
Proven at scale
30M+
health checkups performed
100+
customers & partners globally
20+
languages supported
safety check
Built for clinical safety — not a general-purpose chatbot
Infermedica pairs validated medical knowledge with generative dialogue — neuro-symbolic AI that's natural to talk to and safe to deploy.
General-purpose LLM
- Not trained or governed for clinical triage
- Prone to hallucination and unpredictable outputs
- Black box — limited transparency and auditability
- No medical-device certification or clinical accountability
Infermedica
- Grounded in human-verified medical content
- Deterministic, explainable clinical outputs
- Built & validated by 160,000+ hours of doctor’s input
MGP is CE-marked as Class IIb medical device software under Regulation (EU) 2017/745