🎯 A live demo built by Shanto Mathew for the Retell AI Forward-Deployed Engineer interview. · Source on GitHub · LinkedIn · Email
Copilot is live · +1 (940) 353-0737

Build a production-ready
voice agent in 60 seconds.

Call Copilot. Tell it what you want. A real Retell agent — with a real voice, a real phone number, real tools — gets provisioned live on the call. Or skip the phone and fill out the form. Same backend.

~6s
provisioning latency
1 call
from spec to live agent
Retell-native
LLM + Agent + Voice
Built in 6h
by one engineer
👋 This whole thing — the voice Copilot, the web form, the short-code system, the Retell + AWS backend, the Netlify edge — is a hand-built interview demo for the Retell AI Forward-Deployed Engineer role. Shipped end-to-end in under a day by Shanto Mathew.
Path A

📞 Call the Copilot

Pick up your phone. Tell Copilot what your agent should do, what voice, what to ask callers. It reads back a short URL while you're still on the line.

Answers instantly. Talk like you'd talk to a person. It's fast.

Path B

💻 Build it here

Same backend, same latency, no phone needed. You get a "Test my agent" URL the moment it's live.

How it works

From "describe it" to "call it" in three steps.

1

Describe the agent

By voice or by form, tell Copilot the agent's job, name, voice, greeting, and the questions it should ask.

2

Copilot provisions

A Lambda behind the Copilot tool hits Retell's REST API, creates a real LLM + Agent tuned to your spec, and mints a web-call access token.

3

You test it — instantly

Copilot reads back a short URL like /t/abxjm. Visit it, hit the call button, talk to your brand-new agent in-browser. No signup, no phone number purchase.

Try these pre-built examples

Click, talk, see what a 60-second build looks like.

Under the hood

One engineer, one day, the full stack.

📞 Retell Voice SDK 🎤 ElevenLabs voices (Marissa · Adrian · Kate · Brian) 🛠 GPT-4.1-mini on Retell LLM ☁️ AWS Lambda + HTTP API Gateway 🗂 DynamoDB (short-code store) 🌐 Retell Web Call SDK (for in-browser test)

The Copilot itself is a Retell agent with a custom provision_agent tool. When you describe the agent you want, the tool calls the Lambda, which calls Retell's /create-retell-llm and /create-agent, stores a short-code mapping in DynamoDB, and returns a memorable URL the Copilot reads back to you. Same pipeline serves the web form. Every piece is something a Retell FDE would touch in their first month.

About this demo

Why I built it.

I'm Shanto Mathew — 12+ years of production Python, and for the last 2 years I've been shipping real voice + agentic-AI systems (Retell, OpenAI Realtime, LangGraph, function calling, sub-250 ms latency).

When I saw the Retell AI Forward-Deployed Engineer role I didn't want to send a resume and hope. I wanted to ship the kind of thing an FDE actually ships in the field — a working, production-lean Retell deployment that a customer could start using the moment I left the call.

So I built Retell Copilot: a voice agent (on Retell) whose one tool creates new voice agents (on Retell) on demand. Same developer experience I'd want a Retell customer to have. The whole system — voice prompt, Lambda, API Gateway, DynamoDB short-codes, the web companion, the Netlify edge — is in the public repo below. That's my cover letter.

Get in touch
📧 shanto12@gmail.com
📱 +1 (213) 431-9641
💼 linkedin.com/in/shanto-mathew
📦 github.com/shanto12/retell-copilot
Dallas, TX · Work authorized (E-3D, no sponsorship required)