/ overview
An AI support assistant for Easebuzz merchants, grounded in internal product documentation and payment policy so answers reflect how the platform actually behaves.
Bedrock-hosted models sit behind a Django service that handles retrieval, prompt assembly, guardrails and conversation state, keeping model access server-side.
The assistant absorbs repetitive merchant queries — settlement timelines, onboarding steps, payment status — and escalates anything outside its grounding to a human agent.
/ what it does
- 01Retrieval over internal product and policy documents for grounded answers.
- 02Django service layer owning prompts, guardrails and conversation state.
- 03Escalation path to human support when confidence or scope fails.
- 04Deployed into live merchant support operations.
- 05Model access kept entirely server-side, with no key exposure to clients.
/ numbers
production
environment
merchants
audience
internal docs
grounding
/ stack
AI
Amazon BedrockRetrieval groundingPrompt guardrails
Backend
PythonDjangoREST
Infra
AWSDockerJenkins
proprietary code — built at Easebuzz, not public