CASE STUDY · HEALTHCARE & LIFE SCIENCES
AI clinical decision support for integrative medicine
We built an AI-powered platform that helps Ayurvedic doctors analyse cases, check safety and explore research, with every answer traced back to verified sources, alongside the portals clinics and hospitals need to run day to day.
THE CHALLENGE
What needed to happen
Practitioners of traditional medicine rely on a vast body of classical literature that is slow to search and hard to apply consistently. The client wanted AI that could genuinely support clinical reasoning, but in healthcare an unsupported or unsafe answer is worse than no answer. Every recommendation had to cite its sources, flag interactions with modern medicines, and keep patient data private, all inside a platform that doctors, clinics and hospitals could adopt on their own.
WHAT WE DELIVERED
Evidence-first AI,
built for clinicians.
Clinical AI engine
- Retrieval-augmented reasoning over a curated library of classical texts
- A clinical knowledge graph linking conditions, herbs, actions and formulations
- A multi-step plan, retrieve, reason, validate and reflect loop
- Citation verification so every claim maps to an exact source
- Model routing across AI providers with cost caps and caching
- Golden-dataset evaluation and model drift detection
Doctor portal
- Patient records and case management
- AI case analysis with references the doctor can inspect
- Private case chat grounded in each patient’s record
- Review queue for AI-generated assessments
- Appointments, availability and schedule blocks
- Team management with granular permissions
Research & safety workspace
- Herb–drug interaction checker that runs without AI or patient data
- Repurposing evaluator for existing botanicals, with exact citations
- Formulation hypothesis generator with automatic safety re-checks
- Causality explainer that maps evidence pathways, gaps and conflicts
- Clear labelling of hypotheses versus established evidence
- Local export of results without storing patient records
Patient self-assessment
- An adaptive interview that chooses each next question from previous answers
- A probability engine that builds a structured health report
- Live progress based on real work completed, not timers
- Downloadable PDF reports
- Stateless design: the engine stores no patient data
- Paid assessments with secure payment links and QR codes
Platform & administration
- Self-service onboarding for doctors, clinics and hospitals
- Credential checks, email verification and admin approval
- Subscriptions, pricing, coupons and payment tracking
- Question bank and AI skill registry management
- Per-organisation feature controls and system statistics
- Secure support access through audited impersonation
ENGINEERING & OPERATIONS
- Strict validation of every AI response against a schema before it reaches a doctor
- No fabricated fallbacks: if evidence or safety checks fail, the system says so
- Role- and organisation-scoped access control with rate limiting on AI tools
- An OpenAPI contract covering the full backend
- Streaming responses for long-running AI generation
- Separate development, staging and production knowledge bases with promotion tooling
- Automated test suites across the API and AI services
- Containerised services ready for cloud deployment
WHAT IT MADE POSSIBLE
- Faster, evidence-backed answers from a body of literature that once took hours to search
- Every AI suggestion traceable to its source text
- Safety screening for interactions with modern medicines built into every tool
- Privacy by design, with patient data kept to the minimum needed
- A platform clinics and hospitals can join and run without engineering help
- New revenue through subscriptions and paid patient assessments
TECHNOLOGY
Tools we used
MORE WORK
