terrals

Training data for medical AI

Training data for medical AI: raw signal for pre-training, deep cohorts for fine-tuning, gold answers for SFT, expert votes for RLHF, real outcomes for RL environments, and ground truth for medical AI.

01Pre-training02Fine-tuning03SFT04RLHF05RL

Multi-modal clinical records, radiology, whole-slide pathology and prospective omics — linked at the patient level, de-identified at source, and shaped for every stage of the training stack.

  • WGS✦
  • WES✦
  • NGS panels✦
  • Bulk RNA-seq✦
  • scRNA-seq✦
  • CT✦
  • MRI✦
  • X-ray✦
  • Mammography✦
  • PET-CT✦
  • Whole-slide H&E✦
  • IHC✦
  • EMR✦
  • Labs✦
  • Medications✦
  • Discharge summaries✦
  • Longitudinal follow-up✦
  • RL environments✦

Not another dataset. The ground truth medicine's AI learns from — every layer of every patient, ready for every stage of training.

The training stack

From the first token to the last reward.

Whatever stage your model is in, we supply the clinical signal it needs next.

01Pre-training

Corpus-scale clinical data.

De-identified notes, reports, DICOM, whole-slide images and sequence data — deduplicated, tokenisation-ready and linked across modalities.

  • Clinical text
  • Image patches
  • Sequence k-mers
  • Multi-modal pairs
dyspnoeaonexertionHbA1c8.1%→metforminECOG1biopsyIDCgradeIIdyspnoeaonexertionHbA1c8.1%→metforminECOG1biopsyIDCgradeII
ER+PR+HER2-Ki-6730%mitosesnuclearpleomorphismstromaER+PR+HER2-Ki-6730%mitosesnuclearpleomorphismstroma
ACGTTGGATTCATTAGGCCCGATAAGCTTAGGCATCTACGGACTTAAGACGTTGGATTCATTAGGCCCGATAAGCTTAGGCATCTACGGACTTAAG
your model
02Fine-tuning

Domain adaptation, by specialty.

Curated cohorts for oncology, cardiology, radiology and more — balanced, labelled and split for continued training.

  • Specialty cohorts
  • Report–image pairs
  • Labels & outcomes
general+ specialty datasteps →eval loss
03SFT

Expert-written demonstrations.

Instruction–response pairs authored and reviewed by practising clinicians, grounded in real (de-identified) cases.

  • Clinician-authored
  • Case-grounded
  • Multi-turn
62F, breast lump 2.1 cm, core biopsy: IDC grade II, ER+/PR+/HER2−. What's the next step?
authored by oncologist · reviewed ×2
04RLHF

Clinician preference at scale.

Panels of doctors compare and rank model outputs with rubric-based rationale — the signal reward models need.

  • Pairwise ranking
  • Rubrics
  • Rationales
Response A

Start chemotherapy immediately for all patients.

  • ✗ Safe
  • ✗ Guideline-concordant
  • ✓ Cites evidence
Response B

Surgery first; use genomic risk to decide on chemo.

  • ✓ Safe
  • ✓ Guideline-concordant
  • ✓ Cites evidence
reward model signal…
05RL environments

Rewards from real outcomes.

Environments where the agent's plan is scored against what actually happened next in the patient's record.

  • Outcome-anchored
  • Leakage-checked
  • Deterministic reward
Agentepisode ↻Staterecord to day 0Actionorders · meds · planRewardvs. real day 90

The Terrals record

One patient. Every layer.

Most datasets give you one slice of a person. We link every signal the hospital captured, in the order it happened.

  1. 01Clinical timeline
  2. 02Radiology
  3. 03Pathology
  4. 04Omics
  5. 05Linked & de-identified

Diagnoses, labs, medications, procedures and notes — years of encounters on one axis.

2019202020212022202320242025EMRImagingH&EOmicsHypertensionLipid panelBreast lumpCore biopsyChemoSurgeryFollow-upWSI · 40×core biopsyPIK3CATP53RNA-seq · tumour vs normalpatient · tr_8f3a…c21 · de-identified
Illustrative, synthetic patient. No real patient data is shown on this site.

What we do

Eight ways we work with you.

From a single imaging cohort to a prospective, multi-omic study — or putting your own data and AI to work, privately and DPDP-ready.

02 · Prospective omics

Samples collected for your question — not someone else's.

We recruit, consent and collect prospectively through partner hospitals, then sequence to the depth and platform you specify.

01

NGS panels

Targeted oncology & hereditary panels

02

WGS

30×–60× germline, tumour–normal pairs

03

WES

Exome, tumour–normal

04

Bulk RNA-seq

Tissue & blood transcriptomes

05

scRNA-seq

Single-cell expression from fresh tissue

06

Methylation

Array or sequencing based

Sample journey

  1. 1Consent
  2. 2Collection
  3. 3Sequencing
  4. 4QC
  5. 5Linked delivery
Built to your spec:Cohort & inclusion criteriaTissue & sample typePlatform & depthMatched clinical follow-up
AXIAL · CHESTW 1500 / L −600SLICE 015 / 0281.0 mm

Synthetic, procedurally generated. Scroll to move through the volume.

03 · Imaging

Pixels with context.

Full DICOM series and whole-slide images, each paired with the report and the patient's clinical story around it.

CTMRIX-rayMammographyPET-CTUltrasoundWhole-slide H&EIHC
Format
Native DICOM · SVS / TIFF for WSI
Paired with
Radiology & pathology reports
De-identified
Headers, burned-in text and reports

04 · RL environments

Train clinical agents against what actually happened.

Outcome-anchored environments: the agent sees a real patient's record cut off at a decision point and proposes a plan. The reward comes from that patient's own later record — no human labelling in the loop.

  1. 01

    State

    De-identified record up to a decision point

  2. 02

    Action

    Investigations, medications, follow-up plan

  3. 03

    Reward

    Scored against the patient's real next months

  • ◆Leakage-checked: the agent cannot see the future
  • ◆Deterministic scoring, no LLM judge
  • ◆Thousands of episodes per specialty
episode viewer · illustrative

06 · DPDP compliance

Ready for India's Digital Personal Data Protection Act.

Health data is personal data. We help hospitals, diagnostic chains and health-tech companies meet the DPDP Act 2023 and DPDP Rules — without stopping research or care.

01Readiness assessment

Gap report against the Act and Rules, prioritised by risk.

02Data mapping

What personal data you hold, where it flows, who touches it.

03Consent & notices

Clear notices and consent flows, including Consent Manager integration.

04De-identification

EMR, DICOM imaging and pathology, de-identified for research use.

05Contracts & vendors

Data processing agreements and third-party reviews.

06Breach readiness

Response playbook and notification workflow.

07DPO-as-a-service

Ongoing support and staff training.

Book a DPDP assessment Advisory and implementation support; not legal advice.
Patient recordscanning · 0/4 identifiers
Name
Asha R. (fictional)
Phone
+91 90000 00000
Address
12, Example Road, Pune
MRN
EX-4471-22
Age
62 → 60–64
Diagnosis
Invasive ductal carcinoma
HbA1c
7.2 %

Fictional record for illustration.

Privacy is the product.

01

De-identified at source

Identifiers are removed inside the hospital before data is released.

02

Consented collection

Prospective samples are collected under informed consent.

03

Ethics-approved

Studies run under institutional ethics committee approval, following ICMR guidelines.

04

DPDP-aligned

Processing designed around India's DPDP Act 2023.

05

Secure delivery

Encrypted, access-controlled, time-limited delivery.

Request data

What does your model need?

Tell us the modalities, scale and timeline. We'll come back with a sample and a scoped proposal.

Prefer email? hello@terrals.bio

What data do you need?