HelloMed

HelloMed is an end-to-end hospital management system I designed for clinics that need digital workflows without committing to a heavyweight EMR. It models the encounter, not the patient, so every visit, prescription, and lab order belongs to a specific episode of care. On top of the relational core sits an Ollama-powered recommendation engine that suggests differential diagnoses from symptoms — it suggests, never decides, and every AI suggestion is logged with the prompt and model version.

A quick walkthrough of the encounter workflow:

Curious how the AI suggestions look in practice? Check the GitHub repo↗ for the prompt templates, or read about the Laravel docs↗ on how the routing layer ties it all together.

Highlights

• Relational schema modelling encounters, prescriptions, lab orders, and results with referential integrity
• Ollama-powered differential-diagnosis suggestion engine with full audit log of prompts and model versions
• Role-based access control for doctors, nurses, reception, and admin with route-level guards
• Audit trail: every record update is timestamped and attributable to a user

Challenges

• Designing a schema that scales to multi-clinic deployments without requiring nightly batch reconciliation
• Keeping LLM latency acceptable (< 2s) on small Ollama models while still surfacing a ranked suggestion list

What I learned

• Healthcare software is unforgiving when you can't reconstruct what happened — build the event log first
• AI in clinical settings must suggest, not decide, and every override is part of the audit trail