# Nateeq AI

Case study by Devansh Tiwari (https://www.devanshtiwari.com) · https://www.devanshtiwari.com/work/nateeq-ai

White-label platform for deploying branded AI voice assistants. Multi-tenant SaaS with embeddable widgets.

## Key facts

- Role: Lead Engineer
- Client: OSIT
- Year: 2026
- Status: Shipped
- Tech: NestJS, Next.js, Prisma, PostgreSQL, pgvector, ElevenLabs ConvAI, Turborepo, RBAC

## At a glance

- Problem: A Saudi IT company wanted to resell voice AI, but ElevenLabs only sells an API.
- What I built: A white-label platform where each client gets a branded voice agent, its own data and a dashboard.
- Result: The client can resell voice agents under its own brand, and every call becomes a lead.

## Numbers

| Metric | Value |
| --- | --- |
| Infrastructure to product | 6 weeks |
| Code to embed on any site | One line |
| Separate data store | Per client |

## The problem

ElevenLabs sells a powerful voice AI API. But APIs don't sell to business buyers. A Saudi IT company needed to resell AI voice assistants to GCC enterprises. ElevenLabs gives you agent creation and conversation endpoints. It does not give you multi-tenant isolation, embed code generation, lead extraction, billing, branding, or Arabic support.

## What I built

Took the platform from infrastructure to a shippable product in 6 weeks. Tenants create agents, test with live voice, generate a one-line embed code, deploy on any website, and see extracted leads in a dashboard. Built the ElevenLabs integration layer, widget embedding with cross-domain CORS, AI-powered lead extraction that turns voice conversations into a lead gen tool, and the admin console with RTL. Schema-per-tenant data isolation.

### One-Line Embed

Tenants paste a single script tag into any site. The widget bootstraps itself, negotiates CORS with the tenant's subdomain, and renders a floating voice button. No build step, no SDK, no dependencies.

### Live Voice Agent Builder

Tenants configure agent personality, voice model, greeting script, and fallback behavior in a visual editor. Changes deploy instantly via WebSocket push. Test mode lets you talk to the agent before publishing.

### AI Lead Extraction

After each voice call, an LLM reads the transcript and extracts structured leads: name, phone, email, intent, urgency, budget range. Results appear in a filterable dashboard with CSV export. Turns conversations from a cost center into a revenue driver.

### Schema-Per-Tenant Isolation

Each tenant gets a dedicated PostgreSQL schema with its own tables, indexes, and RLS policies. Cross-tenant queries are physically impossible at the database level. Prisma handles schema switching per request via connection parameter overrides.

## Highlights

- Took the platform from infrastructure to a shippable product in 6 weeks
- One-line embed code with API key auth reduced tenant time-to-deploy from 'hire a developer' to copy-paste
- Built AI lead extraction module that turned voice conversations from a cost center into a revenue driver
- Schema-per-tenant data isolation with cross-domain CORS for subdomain-per-tenant routing

## Work with Devansh

Devansh Tiwari is an AI product engineer based in India, available for freelance and consulting work: custom data feeds from sites without APIs, AI features (agents, chat, RAG, MCP servers) built on that data, and running them after launch.

- Book a 15 minute call: https://cal.com/devansh0718/15min
- Email: devanshtiwari365@gmail.com
- All case studies: https://www.devanshtiwari.com/work
- Site summary for AI tools: https://www.devanshtiwari.com/llms.txt
