NanoSense
Menu
Mode
Read-only
Query modes
Five
Interfaces
REST API and clinician portal

One platform for the care journey.

NanoSense connects your healthcare data to a medical RAG intelligence layer. It retrieves, reasons, and grounds every answer in the patient’s own record. It is read-only: nothing is written back, and your EMR stays the system of record.

How it fits together

Healthcare organizations connect on the left. Patients and caregivers receive grounded insights on the right. NanoSense sits in the middle.

NanoSense Platform Architecture: healthcare organizations connect to the Medical RAG Intelligence API, which shares insights across the patient care circle
Figure 1. Platform architecture.

Query modes

Pick the depth of reasoning your use case needs. Every mode is grounded in the patient’s record and returns a confidence score.

Table 1. The five query modes.
ModeWhat it does
FastFew-shot SQL over the EHR with a grounded summary. Suited to high-volume clinical lookups.
RAG + CAGRetrieval-augmented generation with contextual augmentation. Our recommended mode: balanced accuracy with prompt-caching cost savings.
Deep ThinkingMulti-step LangGraph reasoning for complex clinical questions that need to weigh several parts of the record before answering.
RadiologyImaging-aware queries that pull in studies and radiology reports when a question is about imaging findings.
MCPModel Context Protocol tools for agentic integrations: query the EHR, search the knowledge base, and reach external sources from an agent.

Not sure which? Start with RAG + CAG, the best accuracy-to-cost balance for most clinical workflows. See plans

From question to grounded answer

Every query follows the same auditable path, and the order matters.

  1. AuthenticateA tenant API key or JWT identifies the caller and scopes every result to that tenant’s data.
  2. RetrieveThe question is turned into a validated query over the patient’s conditions, medications, labs, and history.
  3. ReasonThe LLM reasons over the retrieved record with PHI redacted, producing an answer grounded in the record.
  4. ReturnYou get the answer, a confidence score, and the sources used, ready to surface in your app. Nothing is written back.
Figure 2. The path of one query.

Knowledge graph

Beyond flat records, NanoSense links conditions, medications, procedures, and observations into a graph. It surfaces relationships a single query does not see.

  • Cross-entity insights, such as how a condition relates to a medication in the record
  • Visualizations for clinicians and care teams
  • Literature mining to connect the record to current evidence

The assistant cites the record and does not replace clinical judgment.

Technology stack

Table 2. Components by layer.
LayerComponents
APIFastAPI, REST, FHIR R4
ReasoningLangChain, LangGraph
DataPostgreSQL, FAISS, LanceDB
ModelsOllama, vLLM, Groq, OpenAI
InfrastructureAWS ECS Fargate, RDS, KMS
ComplianceBuilt for HIPAA, BAA with every customer, SOC 2 Type II in progress

Access

Subscribe through AWS Marketplace, then bring your data and ask your first question.

Request access Read the docs

References

  1. Request the security whitepaperSent by email until the PDF is hosted
  2. Request the Business Associate AgreementSent by email until the PDF is hosted
  3. Security and data handlingControls, retention, subprocessors and shared responsibility