AlgocyteGPt™ | Bayesian differential diagnosis with traceability

Specialised medical large language model

Guiding and auditing causal mechanistic pathways with traceability at every step.

An LLM trained, tested and guided on the aggregated trends of 150M+ blood result records and clinical profiles.

Florence live demo in AlgocyteGPt on a laptop: a synthetic case presentation ready to run, with previous runs listed below — Illustrative Florence 1.0 live demo using a synthetic case. Decision support only, not a diagnosis.

Specialised reasoning. Any frontier model.

Bayesian differential diagnosis

AlgocyteGPt™ is a specialised large language model based on specialised Bayesian differential diagnosis, independent of the choice of frontier model. Each case is worked through as a ranked set of hypotheses that is updated as new evidence arrives.

Coexists with frontier models

AlgocyteGPt™ guides and audits the causal mechanistic pathways of inference engines for full traceability, preserving data provenance and full clinician control.

Agents

Ada 1.0 + Florence 1.0

Specialised AI Agents with Mechanistic Explainability.

Florence 1.0 and Ada 1.0 are AlgocyteGPt's specialised agentic large language models. Founded on Bayesian inference, they help create ultra-personalised questionnaires from all the multimodal data sources collected across the Algocyte ecosystem, to support differential diagnosis and build a clinical digital twin over time.

A case presentation is worked through the full pipeline and returned as a grounded, source-cited assessment with a ranked differential coded in ICD-10, with phecodes and SNOMED CT on top, for review by the clinician in charge. The approach draws on longitudinal disease-signature research.

Ada and Florence are currently in alpha, being rolled out internally with a focus group for testing purposes, and are subject to regulatory compliance in different jurisdictions. All connections are secure, under ISO 27001 and Cyber Essentials Plus.

Better questions. A continuing diagnostic cycle.

Illustrative treatment check-in for a patient on cancer therapy: pain scale, temperature, fatigue, sleep, nausea, appetite, side effects and mood

Automatically generated, ultra-personalised questionnaires combine medical data and blood results with patient responses, lifestyle and wearable information. They are fine-tuned to investigate diagnostic directions suggested by the available evidence.

Ada and Florence never interact with the patient directly, other than through questions vetted and supervised by clinicians. Clinicians remain in full control: they can tweak, delete or add anything, and use two-way communication to enrich the medical record.

The answers are attached to each blood test result to enrich it and inform the next cycle of tests and questions, in an ongoing cycle of continuous care.

Symptoms, lifestyle and blood factors. Together.

No single reading tells the whole story. AlgocyteGPt™ reasons over what the individual reports, everyday context and blood results in one longitudinal record. Uploaded and scanned medical reports become structured, machine-readable data, with the original documents and available sources retained.

Symptoms and wellbeing

Adaptive check-ins capture symptoms, energy, sleep quality, stress and reported experience. Follow-up draws on medical data, earlier replies and daily context, rather than treating each answer in isolation.

Lifestyle and daily context

Available sleep, activity and heart-rate data add everyday context. Compatible sources include Apple Health, Android Health Connect, Dexcom, Fitbit, Oura, Whoop and others.

Blood factors and records

Whether markers come from Algocyte Proxima or other laboratory partners, clinical reports or uploaded results by the user with full traceability, all gets into the same record and enriched pathway.

Illustrative check-in covering sleep quality, energy, stress, symptoms and routine changes — Illustrative Algocyte HealthHub screens. Connected measurements depend on the source, service and consent.

HealthHub lifestyle view with sleep, activity and resting-heart-rate context — Illustrative Algocyte HealthHub screens. Connected measurements depend on the source, service and consent.

Traceable at every step.

Every step of the reasoning can be followed back to the data, the source and the model that produced it.

Data provenance

Each result, document and response keeps a link to its original source.

Auditable pathways

Causal mechanistic pathways are guided and checked, step by step, rather than left inside a black box.

Model independent

The Bayesian layer works alongside the frontier model of choice, so reasoning stays consistent as models change.

Clinician control

Outputs are decision support. The clinician in charge reviews, edits and decides.

AlgocyteGPt™ provides clinician-facing decision support and does not replace clinical assessment. The 150M+ figure refers to aggregated trends in blood result records with clinical profiles, not to access to individual records. Module-specific intended use, availability and regulatory status apply and differ between jurisdictions.

Regulatory information