livedata-context-engine-case-study

Role: Product Manager (spec + build) Stack: TypeScript, Next.js, PostgreSQL, OpenAI Embeddings, MCP, Prisma Status: Live — widely adopted across the organization

> This is a case study. No proprietary source code, customer data, or internal content is included. See NOTICE.md.

---

LiveData serves 90+ hospitals with a complex product suite spanning surgical scheduling, real-time OR coordination, analytics, and patient flow. The company has deep institutional knowledge — product details, pricing, customer personas, competitive positioning, clinical domain expertise, and operational KPIs — spread across documents, tribal knowledge, and individual heads.

For a lean team, that creates a constant tax: every new employee ramps slowly, every sales rep has to remember which battle card applies, every PM has to re-research domain benchmarks before writing a PRD. The knowledge existed. Getting to it was the problem.

The goal was to make that institutional knowledge instantly accessible to anyone in the org, in natural language, calibrated to their role.

---

I designed and built the entire system. This is an internal product in active daily use across customer success, sales, marketing, and product.

---

Rather than a single generic chatbot, the system is organized into purpose-built engines for each function:

| Engine | Primary Users | Optimized For | |--------|--------------|---------------| | Sales Engine | Account executives | Competitive positioning, objection handling, pricing, customer profiles | | CS Engine | Customer success managers | Product deep-dives, implementation guidance, escalation context | | PM Engine | Product managers | Domain benchmarks, feature context, roadmap alignment | | Marketing Engine | Marketing | Brand voice, product messaging, content guidance |

Each e