All case studies
Industrial · Energy
RAG assistant for a 12k-employee enterprise.
Cut average ticket-resolution time from 28 min to 4 min.
RAGMulti-tenant12k users
Outcomes
By the numbers.
28 → 4 minTicket resolution
94%Eval pass rate
12,000+Monthly users
Challenge
The problem we were brought in to solve.
Frontline operators spent hours per shift searching internal SharePoint, Confluence, and PDF runbooks. Existing search returned keyword matches with no context.
Approach
How we built it.
- Indexed 18,000 internal documents into a hybrid keyword + dense vector store.
- Built a retrieval pipeline with re-ranking and citation-aware answers.
- Wrapped the model in an eval harness with 240 hand-graded golden questions to prevent regressions.
- Shipped a single-page chat UI integrated with the company's SSO.
Stack
The technology we used.
Anthropic Claude
pgvector
Next.js
FastAPI
LangSmith
Related services
Have a similar problem to solve?
Tell us about your project. We respond within one business day.