chrisgeaton/livedata-lumen-case-study
Case study: LiveData's first AI product — natural language OR data queries via MCP
Platform-specific configuration:
{
"mcpServers": {
"livedata-lumen-case-study": {
"command": "npx",
"args": [
"-y",
"livedata-lumen-case-study"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
Role: Product Manager (architecture + POC build) Stack: TypeScript, Express.js, AWS Bedrock (Claude Sonnet), MCP Protocol Status: POC complete — engineering team currently building production version on this architecture
> This is a case study. No proprietary source code or customer data is included. See NOTICE.md.
---
LiveData's analytics product gives hospital OR directors access to detailed surgical workflow data — block utilization, first-case-on-time rates, turnover times, cancellation patterns. The data is valuable. The problem is access: extracting a specific insight requires navigating dashboards, selecting the right filters, and knowing which metric to look at in the first place.
Most OR directors don't have time for that. A surgeon asks a question before a block committee meeting. A charge nurse wants to know how their unit is performing. A CSM needs a number to anchor a customer conversation. What they need is to ask a question and get an answer — not a dashboard.
---
I defined the architecture and built the proof of concept. The engineering team is now building the production version using this architecture as the foundation. Lumen is LiveData's first AI product.
This is an unusual outcome for a PM: I specced the product, proved the concept in working code, and handed the engineering team something they could build on rather than a requirements document they had to interpret.
---
Perioperative staff can ask plain-language questions and receive immediate, data-backed answers:
The core of the system is a Claude Sonnet agent running on AWS Bedrock with ac
Loading reviews...