Zhivko Todorov
ALL CASE STUDIES

CASE 162 · KRAIT · 2025

SEARCHOPENSEARCHINTERNAL DOCSKENDRA

Internal docs that the new engineer can actually find.

An engineering org had documentation across Notion, Backstage, GitHub READMEs, and a dozen old Confluence pages nobody had migrated. Searching for an answer meant guessing where it might live. We deployed Amazon Kendra against all four sources with a unified search interface.

INDUSTRY

Engineering org (internal)

DOMAIN

PLATFORM

DELIVERED

2025

STACK

AMAZON KENDRA·KENDRA CONNECTORS·COGNITO·CLOUDFRONT·LAMBDA

RESULTS

What changed, by the numbers.

TIME-TO-FIND-ANSWER

−68%

NEW-HIRE SURVEY

SOURCES UNIFIED

4

NOTION + GITHUB + BACKSTAGE + CONFLUENCE

RELEVANCE TUNING

KENDRA ML

IMPROVES WITH USE

SUPPORT QUERIES IN #ENG-HELP

−42%

SELF-SERVE

HOW IT WENT

The Slack channel #eng-help was the de facto answer system, because the canonical search across systems didn’t work. New engineers asked questions whose answers were technically documented somewhere — they just couldn’t find them.

Kendra’s pre-built connectors ingested Notion, GitHub, Backstage, and the Confluence archive. A small Lambda fronted the Kendra index with a Cognito-authenticated UI that engineers could open from any browser. Kendra’s relevance ML tunes itself from click-through signals over time.

Time-to-find-an-answer dropped 68% in the new-hire survey three months after rollout. Support queries in #eng-help dropped 42% — the questions still come, but they’re harder ones that genuinely need a human. The first 58% of "where is this documented?" answered itself.

READY WHEN YOU ARE

Let's get your AWS bill (and architecture) in order.

The discovery call is free. You walk away with at least one concrete idea — even if we never work together.

Or email directly →