02CASE STUDY

AI Support Platform

ECHO

ROLE

SOLO BUILD — FULL PLATFORM

TIMELINE

SHIPPED · ACTIVELY MAINTAINED

SCOPE

Embeddable widgetRAG pipelineVoice AIOperator dashboardBilling

STACK

Next.jsConvexGemini RAGVapi VoiceTurborepo
ECHO — screenshot

OVERVIEW

A multi-tenant SaaS: companies embed an AI support widget with a single script tag, upload their docs, and Echo answers customers around the clock — escalating to humans the moment it should.

Echo is three independently deployed apps in one Turborepo — the operator dashboard, the visitor widget, and a zero-dependency Vite loader that injects an organization-scoped iframe from a single script tag with a window.EchoWidget API.

The support agent runs Gemini 2.5 Flash through Convex's agent framework with three autonomous tools: semantic search over the org's namespaced RAG embeddings, escalation on detected frustration, and resolution on confirmed satisfaction. Knowledge bases ingest PDF, CSV and TXT with content-hash deduplication — and answers are grounded exclusively in those documents.

Voice support runs on Vapi with live transcripts and call controls, keys encrypted in AWS Secrets Manager. Operators get a realtime inbox with AI-drafted replies, full visitor context (device, browser, country, session history), and one-click human takeover. Multi-tenancy, billing and plan gating run on Clerk organizations; Sentry covers client, server and edge.

OUTCOMES

3deployable apps, one monorepo
24/7AI answers over chat & voice
RAGgrounded in your knowledge base
No hallucinations: if the answer isn't in your knowledge base, a human gets the ticket.
© 2026 RUSHIKESH PALANDEBACK TO TOP ↑
© 2026 RUSHIKESH PALANDE. ALL RIGHTS RESERVED.