Operational AI
An internal RAG search tool at Dawere, built with Vertex AI, that lets any employee search the company's operational manuals instantly.
Operational AI is an internal RAG (retrieval-augmented generation) tool I built at Dawere, so any employee can search the company's operational procedure manuals in plain language instead of digging through folders and documents. It runs on Vertex AI and Google Cloud, with Google OAuth restricting access to company employees only.
The problem
Operational knowledge lived scattered across documents and folders, so employees either interrupted someone else to ask, or gave up looking. The manuals existed; finding the right answer in them, fast, didn't.
The approach
Built a retrieval-augmented generation pipeline on Vertex AI and Google Cloud: the company's operational manuals are indexed and searchable in plain language, so an employee can ask a question and get an answer grounded in the actual manual content instead of a generic AI guess. Google OAuth gates access to the company's own domain, so only employees can sign in. The current frontend is a straightforward HTML and CSS interface, this is the first version, built to prove the approach and get it into employees' hands quickly.
Highlights
- RAG pipeline on Vertex AI, grounding answers in the company's actual operational manuals instead of generic AI output
- Google OAuth restricts access to the company's own domain, employees only
- Live and in use company-wide as the first version, built to validate the approach fast