Case Studies

Medical Practice Patient Portal + AI Triage

Problem

A multi-location medical practice was drowning in phone calls. Patients called to schedule appointments, request prescription refills, ask basic questions about office hours, and check on lab results. Front desk staff spent 80% of their day on the phone, leaving little time for in-person patient care. Wait times were 10-15 minutes, patient satisfaction was declining, and the practice was considering hiring additional staff just to handle call volume.

Constraints

• HIPAA compliance required for all patient data handling
• Integration with existing EHR (eClinicalWorks) without disrupting workflows
• Budget constraint: solution had to cost less than hiring 2 FTEs
• No downtime allowed during implementation
• Staff needed minimal training on new system

What We Built

We built a secure patient portal with AI-powered triage and automation:

• Patient portal with MFA, appointment scheduling, prescription refill requests, and secure messaging
• AI chatbot trained on practice policies, hours, insurance, and common patient questions
• Automated triage: Simple requests (refills, appointments) routed directly to staff task queues; complex questions routed to appropriate provider with context
• HL7 integration with eClinicalWorks for bidirectional data sync (appointments, prescriptions, lab results)
• SMS notifications for appointment reminders and lab result availability

Technologies

• React/TypeScript frontend with responsive mobile design
• Node.js/Express backend with JWT authentication and MFA
• OpenAI GPT-4 for chatbot with RAG (retrieval-augmented generation) over practice documentation
• Mirth Connect interface engine for HL7 ADT/ORM/ORU messages
• PostgreSQL with encrypted patient data fields
• Twilio for SMS notifications

Multi-Location Inventory + Barcode Workflow

Problem

A regional distribution company with 5 warehouses was managing inventory in spreadsheets. Each location had its own Excel file. Transfers between locations required manual data entry at both the sending and receiving warehouse. Inventory counts were inaccurate, leading to stockouts and overstock situations. Picking errors at warehouses resulted in incorrect shipments and customer complaints. The company was losing $50k+ annually to inventory discrepancies and fulfillment errors.

Constraints

• Existing barcode scanners and label printers at each location (no budget for new hardware)
• Warehouse staff had limited computer literacy; solution had to be simple and fast
• No internet connectivity in some warehouse areas; solution needed offline capability
• Integration with QuickBooks for financial reporting
• Phased rollout: one location at a time to minimize disruption

What We Built

We built a centralized inventory management system with barcode-driven workflows:

• Web-based inventory management dashboard with real-time visibility across all locations
• Barcode scanning workflows for receiving, picking, shipping, and transfers
• Offline-capable mobile web app for warehouse floor (syncs when connection restored)
• Automated label printing for incoming inventory and outbound shipments
• Transfer workflows with dual confirmation (scan out at source, scan in at destination)
• Low stock alerts and automated reorder point notifications
• QuickBooks integration for inventory valuation and cost of goods sold

Technologies

• React/TypeScript Progressive Web App (PWA) with offline support
• Python/FastAPI backend with PostgreSQL database
• Zebra printer API for label generation (ZPL format)
• Barcode scanning via web app camera and dedicated USB scanners
• QuickBooks Online API for financial data sync
• Docker deployment on on-prem server with remote VPN access

Legacy System Modernization + Self-Hosting

Problem

A healthcare organization was running a critical clinical workflow application on a Windows Server 2008 physical server with SQL Server 2008. The application was custom-built 15 years ago and had no vendor support. The server hardware was failing (disk errors, overheating), and the organization faced a disaster scenario: no backups were running successfully, and the application would not run on newer Windows versions. They were quoted $250k for a cloud migration by a consultancy, but wanted to keep the system on-premises due to compliance concerns and internet reliability issues at their rural locations.

Constraints

• Zero downtime allowed: application is used 24/7 for patient care
• No application source code available; vendor out of business
• Application must run on Windows Server 2008 / SQL Server 2008 (compatibility issues with newer versions)
• Limited budget: $30k maximum for hardware and implementation
• Must maintain HIPAA compliance

What We Built

We modernized the infrastructure while preserving the legacy application:

• Proxmox VE virtualization cluster with 2 nodes for high availability
• Migrated legacy application to a Windows Server 2008 VM (frozen configuration, isolated network)
• Implemented automated nightly backups with off-site replication
• Set up monitoring and alerting for VM health, disk usage, and application uptime
• Created disaster recovery runbook and tested failover procedures
• Configured firewall rules and VLANs for network segmentation (HIPAA compliance)
• Documented entire infrastructure for future maintenance

Technologies

• Proxmox VE 8.0 for virtualization and HA clustering
• Ceph for distributed storage with replication
• Windows Server 2008 R2 VM (legacy application preserved as-is)
• SQL Server 2008 R2 (database preserved with automated backups)
• PBS (Proxmox Backup Server) for VM-level backups
• Prometheus + Grafana for monitoring and alerting