Long-term medical
intelligence for every patient

We turn scattered reports into one continuous, patient-level disease course — so patients take part in their own care, clinicians understand a new patient in minutes, and researchers work with real-world data that actually spans time.

One patient · 5 hospitals · 2 years · dozens of documents

Partner with us → See the platform →

What we do

An oncology data and
intelligence platform

The next generation of clinical AI is not about reading reports better. It is about putting pathology, imaging, molecular results, treatment, response and out-of-hospital change into a single patient context, and then answering:

  • What has happened so far?
  • Where does this patient stand now?
  • What deserves attention next?

Product

Agents for three roles

Helping patients understand themselves, clinicians understand a patient, and researchers understand a population.

Service

Real-world research delivery

Cohort building and study support for pharma and research institutions. We run studies inside China and deliver analysis and real-world evidence (RWE) — not raw patient data.

Asset

Patient-level longitudinal records

Cross-hospital, continuous, structured and traceable real-world data (RWD), contributed by patients themselves.

Patient base & MVP beta resultsAs of August 2026

1M+

Patient community

Patient-side acquisition — no hospital system integration required

Breadth

30+Tumour types
50+Source institutions

Depth

1,500+Longitudinal patients
30,000+Structured records

Network

1,000+Oncologists in our network
80+Pharma & institutional collaborations

Product · PatientLive

Aji Health

Patient Agent · a records keeper for patients and families

Patients hand over scattered lab reports, discharge summaries and medication records, and get back a readable timeline of their own disease course, plain-language explanations of each report, and follow-up and medication reminders keyed to where they are in treatment.

  • FilingCross-hospital documents in one place, auto-classified and de-duplicated
  • ReadingStaging, lab values and response assessments in language patients understand
  • TimelineTreatment lines, regimen changes and key events reconstructed automatically
  • Follow-upRe-exam reminders, medication logs, symptoms and patient-reported outcomes (PRO)
Visit Aji Health ↗
阿吉健康：档案与首页界面

Products · Clinical & research

Department heads · PIs · Research nurses · Data management

Hospitals & investigators

Clinical Agent Open for co-development

We are refining pre-visit summaries and follow-up prompts with the medical oncology departments of two tertiary hospitals. If your department has a similar need, we would like to hear from you.

  • Pre-visit summaries that cut the time spent assembling a complex history
  • Alerts on out-of-hospital trends, overdue re-exams and adherence gaps
  • Disease-specific cohorts and long-term follow-up, with the out-of-hospital leg carried by the patient app
  • Data preparation and follow-up collection for investigator-initiated trials (IIT)
Start a department conversation →

Medical affairs · HEOR · RWE · Clinical development

Pharma & CRO

Research Agent Open for research partnership

Cohort building and feasibility analysis can already run on the existing dataset. We are looking for our first research partners to define delivery standards and field specifications together.

  • Feasibility analysis and target patient pool assessment
  • Cohort-as-a-Service and external control arm construction
  • Real-world studies, natural history studies and HEOR deliverables
  • Long-term follow-up and outcome capture for post-marketing studies
Request capability deck and sample deliverables →

Agent architectureMemory · Context · Agent

An agent that learns the patient

A general model has enough medical knowledge. What it lacks is this person's context: which line of therapy they are on, how the latest scan compares with the last one, which concepts have already been explained. On top of the model we built three layers — memory, context, and self-improvement.

Patient Memory

Patient memory

A patient state that moves with the disease course. Every interaction is written back, and the next conversation starts from the updated state.

  • Course state: current line of therapy, latest assessment, pending re-exams
  • Comprehension: concepts already explained, questions still open
  • Care context: whether the patient or a family member is using it

Seconds · learning that never touches model weights

Context Engineering

Course context

Two years, five hospitals, dozens of documents — far beyond any context window. The system has to decide what is relevant to this question, right now.

  • Retrieval scoped by disease stage and question type
  • Conflicting cross-hospital records resolved and flagged
  • Citation anchoring — every answer traces back to a specific document

Retrieval and parsing share one course-event schema

Self-improving agent architecture

Four learning loops nested by time scale. Each changes something different, and each is released differently.

Loop

What it changes

Release gate

ArchitectureQuarterly

System structure, capability boundaries, and new tasks.

Human decision

ModelMonthly

Post-training of task-level small models on reviewed samples, staged release after regression tests.

Human review

PolicyWeekly

Automated tuning of prompts, context rules, retrieval strategy and tool orchestration.

Eval-set gate

MemorySeconds

Patient state and memory. Written on interaction, touching no weights or rules.

Fully automatic

Data compliance & qualityPrivacy policy & data statement →

Data security

Built to China's MLPS Level 3 requirements: encryption at rest, least-privilege access, and auditable logs for key operations.

In-China processing, de-identified output

Patient data is stored and processed inside China and does not leave the country. We partner as an in-China research collaborator: what crosses a border is de-identified analysis and evidence, never the raw dataset.

Medical quality control

Three-tier QC: AI extraction, review by medical specialists, and expert spot-checks.

A clear product boundary

The patient app is a records-management and comprehension tool. It does not issue diagnoses or treatment advice, and directs users to in-person care when it detects a high-risk situation.

Explanations and reminders in Aji Health are AI-generated. They do not constitute a diagnosis or treatment advice and do not replace a clinician's professional judgement.

Partner with us

Start with one concrete question

Patients & families

Get Aji Health and start organising your records

Open the mini-program →

Hospitals & investigators

Co-development programme and IIT data support

Send us your requirements →

Pharma & CRO

Capability deck, sample deliverables, feasibility review

Book a capability briefing →