Healthcare Innovation & Hackathon Platform
We designed, engineered, and shipped a multi-tenant innovation and evaluation platform for Qurofy, replacing manual spreadsheets and fragmented forms with an automated, agent-assisted workflow engine.
The Challenge
Healthcare innovation programs were previously managed across disconnected spreadsheets, Google Forms, and email threads. As programs scaled to thousands of participants and hundreds of cross-functional teams, operations broke down.
Fragmented Data & Lack of Single Source of Truth
Participant profiles, squad rosters, and project submissions were stored in disparate sheets, making real-time tracking and auditability impossible.
Multi-Tenant Isolation Requirements
Different healthcare enterprises, hospital groups, and academic institutions needed completely isolated environments, branding, and registration workflows.
Complex Role-Based Permissions
Four distinct actors (Organizers, Participants, Clinical Mentors, Judges) required strictly scoped capabilities and data visibility boundaries.
Manual Multi-Round Evaluation Inefficiencies
Scoring required manual calculation of weighted criteria, resulting in human error, slow turnaround, and zero automated ranking consensus.
The Innovation Lifecycle Architecture
We mapped the end-to-end program lifecycle into an explicit, state-machine driven domain model.
Production System Architecture
Modular separation of web client, backend gateway, stateful LangGraph agent swarms, and high-throughput data layer.
Key Engineering Decisions
Strict Multi-Tenant Isolation
Organizations own programs, users, and telemetry. Every query enforces tenant tenancy constraints at the database and middleware layers to prevent cross-organization data leakage.
Granular Role-Based Access Control (RBAC)
Permissions are scoped across Platform, Organization, and Event levels. Judges only see assigned projects, while Mentors only access booked squad channels.
Dynamic No-Code Registration Schemas
Rather than hardcoding intake questions, organizers configure fields stored in structured JSONB schemas, allowing full question customization per hackathon without code redeployment.
Deterministic Multi-Round Weighted Scoring
Implemented a mathematical normalization engine: score = sum((criterion_score / max_score) * weight). Leaderboards are derived dynamically from judge consensus rather than manually edited tables.
AI & Intelligent Workflows in Qurofy
Every AI capability was built as a reliable tool-calling workflow with strict guardrails and human approval loops.
Program Creation Copilot
Assists organizers in drafting clinical problem statements, defining challenge tracks, and suggesting balanced evaluation criteria based on past successful programs.
Reviewer Copilot
Extracts key technical claims from project repositories and matches them against judging rubrics to prepare briefing notes for expert medical reviewers.
Squad & Skill Matchmaking
Analyzes participant profiles and suggests complementary pairings (e.g. matching clinical doctors with backend engineers and UI designers).
What should your AI system actually be able to do?
Tell us about the workflow. We'll help determine what should be automated, what should remain human-controlled, and what it takes to make the system reliable.