Clinician-prescribed · AI-guided · Supportive care

Nutrition that helps patients tolerate treatment.

OncoNourish turns an oncology patient's clinical profile into a personalised nutrition chart and day-by-day Indian meal plans — built on auditable clinical rules, reviewed and approved by their care team.

Supportive care only — not a substitute for medical advice. No treat / cure / arrest claims.
Every rule shows "why"
ESPEN · ASCO · ICMR
AK
Anita K.
Breast · Stage II · AC-T
Approved
1.4 g/kg
Protein
1650 kcal
Energy
2.2 L
Fluids
Moong dalPaneerCurd (lactose-low)RagiBanana
Raw saladsStreet foodGrapefruit
Clinician-approved
gate before patient
Grounded in established guidance
ESPEN oncology ASCO supportive care ICMR dietary norms
The positioning

OncoNourish keeps patients nourished and strong enough to get through their treatment — it does not treat, cure, or arrest the disease.

What it does

Maintains nutritional status, manages treatment side-effects through diet, and supports protein & calorie targets during therapy.

Who stays in control

The oncologist or dietitian reviews and approves every plan. The patient receives a read-only, care-team-sanctioned chart.

What it will never claim

No statement that food treats, cures, shrinks, or arrests cancer. Supportive-care framing appears on every screen and every export.

The flow

From clinical profile to an approved plan the patient can follow.

The clinician drives the whole flow. The patient view is a shared, read-only page — no medical decision is ever made without the care team.

01
Clinician

Create & intake

Add a patient and fill the structured intake — cancer type & stage, regimen, comorbidities, diet pattern and intolerances.

02
Engine

Generate

The rules engine sets targets and favour / avoid lists; the LLM turns them into clear, affordable, regional meal plans.

03
Clinician

Review & approve

Inspect every recommendation with its source rule, edit if needed, then approve. Nothing reaches the patient unapproved.

Approval gate
04
Patient

Follow & print

The patient opens their approved chart and daily meals — or takes home a clean, printable PDF the clinician hands over.

Personalisation profile

Six inputs that drive every recommendation.

The intake form is the single source of truth. Each field maps to specific rules — change a field, and the plan changes with it.

Disease & stage

Cancer type, stage and disease condition set the clinical baseline for the whole chart.

typestagecondition

Treatment & meds

Regimen, line of therapy and current / concomitant medications drive drug-nutrient cautions.

regimenlinemeds

Comorbidities

Diabetes, renal, cardiac or hepatic conditions adjust targets and constrain food lists.

diabetesrenalcardiac

Anthropometrics

Age, sex, height and weight compute BMI and personalise protein & calorie targets.

age/sexBMItargets

Diet pattern

Veg / non-veg and regional cuisine keep meals familiar, affordable and culturally relevant.

veg / non-vegregion

Intolerances

Lactose, gluten and a standard checklist filter the food and recipe library automatically.

lactosegluten+ checklist
Phase-1 scope Focused on a small set of high-volume cancers — breast, colorectal, head & neck, and upper-GI — with a data model designed so phase-aware regeneration can be added later.
Engine architecture

Auditable rules first. The LLM only localises.

Clinical decisions live in editable, evidence-based rules. The language model presents and translates them into plain, regional meal plans — it never invents clinical claims or overrides a rule.

Layer 1 · Rules engine

Evidence, encoded

ESPEN / ASCO / ICMR guidance stored as explicit, auditable config — favour & avoid lists, protein & calorie targets, hydration, supplement flags and drug-nutrient cautions, keyed off the profile. Editable by a clinical advisor, not hardcoded.

# protein_target.rule.yaml
when: cancer_stage in ["II","III"]
and: comorbidity != "renal"
set:
  protein_g_per_kg: 1.4
  source: "ESPEN 2021 §4.2"
  why: "preserve lean mass in therapy"
Layer 2 · LLM presenter

Plain, local, affordable

Takes the rules-engine output as structured context and turns it into clear, culturally relevant Indian meal plans in plain language — abstracted behind a single service module with one prompt template.

  • Localises to region, cuisine and budget
  • Respects every intolerance & avoid rule
  • Never adds a clinical claim of its own
Profile Rule Recommendation Meal plan
Full traceability. Every recommendation links back to the rule that produced it, so the clinician can always see why — and audit or edit the source.
Two outputs, one handover

A nutrition chart and daily meal plans.

Generated together, reviewed together, and exported to a single clean, printable PDF the clinician hands to the patient.

Nutrition Chart PDF
1.4 g/kg
Protein
1650
kcal / day
2.2 L
Hydration
Favour
  • High-protein dals & paneer
  • Soft-cooked vegetables
  • Ragi, banana, curd
  • Small frequent meals
Avoid
  • Raw / street food
  • Grapefruit (drug interaction)
  • Very spicy or acidic food
  • Unpasteurised dairy

Nutrition Chart

The framework: favour / avoid lists, macro & protein targets, hydration, supplement guidance and do's & don'ts — mapped to regimen and comorbidities.

Daily Meal Plan · Day 1 Veg · South Indian
7:30 AM
Ragi idli + moong sambar

Soft, easy to swallow · high protein

320 kcal
10:30 AM
Banana + soaked almonds

Energy-dense snack between meals

180 kcal
1:00 PM
Curd rice + paneer bhurji

Lactose-low curd · protein-forward

450 kcal
4:30 PM
Moong dal chilla

Light, protein-rich evening meal

240 kcal
8:00 PM
Khichdi + steamed veg

Gentle on the gut · well tolerated

380 kcal

Daily Meal Plans

Concrete day-by-day meals, portions and simple recipes — built around local food and the patient's exact intolerances.

Safety & compliance

Built as supportive care, with guardrails that don't bend.

Health data is handled carefully and every plan passes through the care team. These are hard requirements, not options.

Clinician approval gate

No plan reaches a patient until the oncologist or dietitian has reviewed and approved it.

Supportive-care framing

A visible disclaimer on every screen and export — never a claim to treat, cure, or arrest disease.

Minimal, consented data

No unnecessary PII, a clear consent notice, and basic access control separating clinician and patient views.

Drug-nutrient cautions

Medication-aware flags surface foods to avoid for known interactions — each traced to its source rule.

Ready when you are

Keep patients nourished
through every cycle.

See OncoNourish run end-to-end on real sample cases — clinician intake, generate, review, approve and PDF handover.