PrecepGo - Agent Dashboard

PrecepGo automates CRNA education tasks using Google ADK agents: clinical scenarios, evaluations, safety monitoring, and compliance tracking. What takes hours now takes minutes.

Comment

PrecepGo - Agent Dashboard

Try Yourself: https://precepgo-agents-frontend-724021185717.us-central1.run.app/
Medium Article: https://medium.com/@bytebauble/from-paper-evaluations-to-ai-agents-building-precepgo-a1ee4c1b7a43
X (Twitter): https://x.com/Josh_Burleson/status/1987958689458966688?s=20

Logo

Overview

PrecepGo Agent Dashboard is an AI-powered automation platform designed specifically for CRNA (Certified Registered Nurse Anesthetist) education programs. Built with Google's Agent Development Kit (ADK), the platform transforms time-intensive educational tasks into automated workflows, allowing faculty to focus on what matters most: teaching and mentoring students.

Inspiration

After graduating from Mayo Clinic, my wife became Assistant Director for a prominent DNAP program on the West Coast. In 2023, we built PrecepGo to modernize clinical evaluations using Google Cloud, successfully deploying cloud-based evaluations across the West Coast.

However, we knew there was more we could do for students, preceptors, and administrators. When the Google ADK Hackathon was announced, I saw the opportunity to evolve PrecepGo into an intelligent agent-powered platform that could truly transform CRNA education through automation.

The Problem

CRNA program directors and faculty face overwhelming administrative burdens:

  • Hours spent creating clinical scenarios for student practice
  • Manual evaluation processing with repetitive documentation
  • Difficult safety monitoring across hundreds of student evaluations
  • Complex COA compliance tracking requiring constant attention
  • Time-consuming site and preceptor analytics

These tasks consume valuable time that should be spent on direct student education and mentorship.

Our Solution

PrecepGo automates these workflows using intelligent AI agents that work together as a coordinated system. Each agent specializes in a specific educational task, processing data from Firestore, generating insights with Gemini AI, and creating actionable outputs.

Key Features

🎯 Clinical Scenario Generation

Automatically creates personalized clinical case studies tailored to individual student needs. Scenarios include:

  • Patient demographics and medical history
  • Anesthesia considerations and challenges
  • Decision points with multiple pathways
  • AI-generated medical images using Imagen

📊 Automated Evaluations

Generates comprehensive student evaluations with:

  • Structured scoring across 24+ clinical competencies
  • Detailed narrative feedback
  • Automatic COA standard mapping
  • Firestore integration for permanent records

🚨 Safety Monitoring

Proactively identifies concerning student performance:

  • Scans evaluations for dangerous ratings (-1 scores)
  • Generates notification records for program directors
  • Tracks patterns across multiple evaluations
  • Enables rapid intervention when needed

📋 COA Compliance Tracking

Maintains accreditation standards automatically:

  • Maps evaluation metrics to COA Standard D requirements
  • Aggregates student performance data
  • Generates compliance reports for audits
  • Tracks progress toward certification milestones

🏥 Site Analytics

Analyzes clinical placement effectiveness:

  • Identifies high-performing preceptors
  • Tracks case type distribution across sites
  • Generates insights on student experiences
  • Supports data-driven placement decisions

⏱️ Time Savings Analytics

Quantifies platform impact with real metrics:

  • Calculates time saved per task type
  • Provides ROI analysis for stakeholders
  • Tracks usage across different timeframes
  • Demonstrates value to administrators

Technology Stack

AI & Machine Learning

  • Google Gemini AI - Natural language processing and content generation
  • Google ADK (Agent Development Kit) - Multi-agent orchestration
  • Vertex AI Imagen - Medical image generation for scenarios

Backend & Data

  • FastAPI - Modern Python web framework
  • Google Cloud Firestore - NoSQL database for evaluations and state
  • Google Cloud Storage - Image and asset storage
  • Python 3.11 - Core application language

Deployment & Infrastructure

  • Google Cloud Run - Serverless container deployment
  • Docker - Containerization
  • Cloud Build - CI/CD pipeline

AI Agent Architecture

  • Root Agent - Main coordinator and entry point
  • Safety Pipeline - Sequential workflow (Evaluation → Notification → Scenario)
  • Evaluation Agent - 7-step sequential evaluation creation
  • Scenario Agent - 6-step clinical scenario generation
  • Notification Agent - 3-step safety monitoring
  • COA Agent - Compliance tracking and reporting
  • Site Agent - Analytics and insights
  • Time Agent - Usage metrics and ROI calculation
  • Image Generator - Medical illustration creation

How It Works

  1. Data Input: Program data (students, cases, templates) loaded from JSON and Firestore
  2. Agent Orchestration: Root agent coordinates specialized agents based on task type
  3. AI Processing: Agents use Gemini models to analyze, generate, and structure content
  4. State Management: Shared state enables agents to pass data in multi-step workflows
  5. Output Generation: Results saved to Firestore with metadata and timestamps
  6. Dashboard Access: Faculty interact with agents through web-based control panel

Real-World Impact

Time Savings

  • Scenario creation: 45 minutes → 3 minutes (93% reduction)
  • Evaluation generation: 30 minutes → 2 minutes (93% reduction)
  • Safety audits: 2 hours → 5 minutes (96% reduction)
  • Compliance reports: 4 hours → 10 minutes (96% reduction)

Scale

  • Processes hundreds of evaluations automatically
  • Generates dozens of clinical scenarios on demand
  • Monitors thousands of data points for safety concerns
  • Supports multiple CRNA programs simultaneously

Quality

  • Maintains COA accreditation standards automatically
  • Provides consistent, structured feedback across all students
  • Enables data-driven decision making for program directors
  • Ensures immediate response to safety concerns

Getting Started

Prerequisites

  • Google Cloud Project with billing enabled
  • Firestore database configured
  • Cloud Storage bucket for images
  • Gemini API key

Quick Deploy

# Set environment variables
export PROJECT_ID=your-project-id
export GOOGLE_API_KEY=your-api-key

# Deploy to Cloud Run
gcloud run deploy precepgo-adk-panel \
  --image gcr.io/$PROJECT_ID/precepgo-adk-panel \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars="GOOGLE_API_KEY=$GOOGLE_API_KEY,FIREBASE_PROJECT_ID=$PROJECT_ID"

Access Dashboard

Navigate to your Cloud Run URL + /dashboard to access the control panel.

Project Structure

precepgo-adk-panel/
├── agents/              # AI agent implementations
│   ├── root_agent.py    # Main coordinator
│   ├── scenario_agent.py
│   ├── evaluations_agent.py
│   ├── notification_agent.py
│   ├── coa_agent.py
│   ├── site_agent.py
│   ├── time_agent.py
│   └── image_agent.py
├── main.py              # FastAPI application
├── data/                # JSON data files
├── context/             # Documentation
└── Dockerfile           # Container configuration

Built For Educators

PrecepGo was created by educators who understand the challenges of CRNA program administration. Every feature addresses a real pain point identified through direct faculty feedback and program director insights.

Deployment Status

✅ Production Ready - Deployed on Google Cloud Run
✅ Scalable - Handles concurrent requests and multiple programs
✅ Secure - Firestore rules, IAM permissions, and Cloud Run authentication
✅ Monitored - Cloud Logging and error tracking enabled

Support & Documentation

  • Dashboard: /dashboard - Interactive control panel
  • API Docs: /docs - FastAPI auto-generated documentation
  • Health Check: /health - Service status endpoint

License

Built for the Google ADK Hackathon 2025.

Empowering CRNA educators through intelligent automation.

Built With

Share this project:

Updates

Submission history