Evarx Standard LLM
Hosted proprietary general-purpose LLM. Ideal for prototyping, low-sensitivity workloads, and high-volume tasks.
- Hosted by Evarx · in-region
- Pay-per-token
- Available across India region
- Quickest path to first agent
The platform
One platform from raw clinical data to a deployed medical agent.
Evarx is composed of four interlocking layers. Each is independently swappable, so you can start hosted and graduate to a fully private fine-tuned SLM without rewriting a workflow.
- 01 · Intelligence engines
- 02 · Data layer & RAG
- 03 · Workflow builder
- 04 · Deploy & scale
01 · Intelligence
Three engines, one orchestrator.
Workflows reference an engine by alias. Swap underlying models — Evarx Standard, Evarx Medical SLM, or your fine-tune — without changing a single line of agent code.
Evarx Medical SLM
A small language model pre-trained on biomedical and clinical corpora. Deployed in your VPC or on-prem.
- Pre-trained on medical corpora
- On-prem or VPC
- Zero data egress
- DPDP Act + HIPAA-aligned
Rapid Fine-Tuned SLM
Upload your data, get a CPU-runnable model trained on it in hours. No ML team required.
- <6 hour fine-tune cycle
- CPU-runnable, quantised
- Air-gapped deploy
- Continuously refined
| Capability | Standard | Private | Custom (USP) |
|---|---|---|---|
| Data residency | Evarx · in-region | Your VPC / on-prem | Your VPC / on-prem |
| Domain knowledge | General | Medical baseline | Your proprietary data |
| Hardware | Hosted | GPU / CPU | CPU |
| Time to deploy | Minutes | Hours | Hours (post fine-tune) |
| Continuous improvement | — | Optional | Built-in |
| Air-gapped support | — | Yes | Yes |
02 · Data layer
Your data, retrieved and refined.
Connectors feed structured and unstructured medical data into a hybrid retrieval index. The same data also generates training pairs to refine your private SLM over time.
Connectors · 30+
- Epic & Cerner EMRClinical
- PubMed / MEDLINELiterature
- ClinicalTrials.govTrials
- OpenFDA / CDSCORegulatory
- SharePointDocuments
- ConfluenceDocuments
- S3 / Azure BlobStorage
- Postgres / MSSQLDatabases
- Snowflake / BigQueryWarehouse
- Veeva VaultLife sciences
- Salesforce Health CloudCRM
- REST / GraphQL APIsCustom
Pipeline
- 01
Ingest
Streaming + batch, with PHI redaction at the edge.
- 02
Chunk + embed
Medical-aware chunking. BGE-M3 multilingual embeddings.
- 03
Index
Hybrid (vector + lexical) over Qdrant + Postgres.
- 04
Retrieve
Re-ranking with grounded citations on every span.
- 05
Refine
Validated outputs become supervised pairs for the next fine-tune.
03 · Workflows
Drag, drop, ship.
Build agents visually or start from medical templates. Every node is observable, versioned, and runs against any of your engines.
Clinical trial protocol summarizer
ICSR triage & MedDRA coding
Pre-visit EHR summarization
Promotional content compliance review
Medical literature daily digest
Multi-step regulatory submission drafter
Plus six more in the agents gallery. All templates are forkable.
04 · Deploy & scale
One workflow. Three deployment shapes.
Choose where each workflow runs. Switch shapes per environment without rewriting agent logic.
Managed SaaS
Hosted by Evarx in AWS Mumbai. Fastest start, single-tenant database.
Customer VPC
Deployed via Terraform into your AWS, Azure, or GCP account. Zero egress.
On-prem · CPU
Docker images for air-gapped hospital networks. Quantised SLMs run on existing servers.
API + SDK
OpenAI-compatible endpoints, TS + Python SDKs.
Audit & RBAC
Per-tenant logs, fine-grained roles, SSO.
Usage billing
Per-call metering, budget caps, finance-ready exports.