Cover Image for Vibe-Code a Voice AI Agent in 3 Hours — Live Workshop with Claude Code - Open Registration
Cover Image for Vibe-Code a Voice AI Agent in 3 Hours — Live Workshop with Claude Code - Open Registration
Avatar for Frontier Tower SF
Presented by
Frontier Tower SF
A 16-floor nexus for frontier tech in SF. Join us at https://frontiertower.io/apply
3 Going

Vibe-Code a Voice AI Agent in 3 Hours — Live Workshop with Claude Code - Open Registration

Registration
Welcome! To join the event, please register below.
About Event

You don't need to write a voice agent from scratch. You need to take a working one, reshape it into your product with Claude Code, deploy it live, and learn how the best teams stress-test and fix voice AI in production.

By 7:30 PM Thursday, you'll have a custom voice AI agent running on LiveKit Cloud — your idea, your prompt, your tools — battle-tested against 50 synthetic callers and debugged with AI. Pizza and drinks on us. 🍕🍻

How it works We start with a fully working LiveKit voice agent on the fastest open stack: Groq + Deepgram + ElevenLabs, sub-500ms latency, Noveum tracing pre-wired. You fork it, then use Claude Code to reshape it into whatever you want to build — a pizza bot, healthcare triage, sales SDR, trivia host, customer support agent for your actual SaaS. Whatever you bring, Claude ships it with you. Everything happens in two terminals — no browser required. Claude Code drives the Noveum MCP to set up endpoints, generate personas + scenarios, run synthetic batches, fetch reports, and apply fixes. Zero clicks.

The voice stack (pre-wired, you don't configure it)
- LiveKit Cloud — real-time voice infra + free deploys
- Deepgram Nova-2 — streaming speech-to-text (~150ms)
- Groq gpt-oss-120b — sub-200ms LLM inference
- ElevenLabs Turbo v2.5 — natural, low-latency TTS
- Noveum — tracing + NovaSynth synthetic users + NovaPilot AI debugging

Total turn latency target: under 500ms — the threshold where voice agents start feeling fluent, not robotic.

Please clone this repo and follow the instructions before the event ->

https://github.com/Noveum/voice-agent-workshop-starter

Agenda (3 hours)

- 4:30 – 4:45 — Welcome, intros, why voice AI breaks in production (pizza is ready 🍕)
- 4:45 – 5:15 — 🛠 FORK & DEPLOY — clone the starter, push to LiveKit Cloud
- 5:15 – 5:45 — 🎨 CUSTOMIZE WITH CLAUDE — one prompt reshapes the agent into your product
- 5:45 – 6:30 — 🎭 STRESS-TEST — Claude drives Noveum's MCP: 5 personas, 5 scenarios, a 10-call batch
- 6:30 – 7:00 — 🔍 DEBUG — traces show exactly where ASR, LLM, or TTS broke
- 7:00 – 7:20 — 🤖 FIX — Claude applies AI-suggested fixes, redeploys, and re-runs the batch
- 7:20 – 7:30 — Demos, Q&A, more pizza, networking 🍻

Pre-work (do this before Thursday — about 25 minutes)

We have 3 hours, and we want to spend them on the build → test → fix loop, not installing things.

- Install Claude Code (or Cursor) — npm install -g @anthropic-ai/claude-code or download Cursor at cursor.com
- System dependencies — Python 3.10+, Node 20+, git, the LiveKit CLI (brew install livekit-cli on macOS)
- Free accounts + API keys (total cost: $0) — Anthropic, LiveKit Cloud, Deepgram, Groq, ElevenLabs, Noveum. All have free tiers covering the workshop.
- Fork the workshop starter — repo link goes out 24 hours before the event. Run the included verify_setup.py script — green checks mean you're ready.
- Bring a voice agent idea. Anything — silly or serious. Best workshops we've run had a Pokémon trivia bot, a yoga instructor, and a fintech compliance auditor in the same room.

Stuck on setup? Show up 30 minutes early on Thursday — we'll have a setup-help corner from 4:00 PM.

What's provided 🍕🍻
- Pizza — fresh and hot at 4:30, more arrives mid-workshop
- Drinks — sparkling water, soda
- Snacks for the long stretches - Power strips at every workspace
- Strong WiFi — but we'll also share a backup hotspot just in case

Who this is for
Founders, engineers, PMs, and designers building or planning tobuild voice AI products — customer support bots, voice IVR, sales agents, healthcare assistants, consumer voice apps. No senior-engineer credentials required. If you can prompt Claude Code, you can ship today.

Hosted by

Shashank Agarwal — building Noveum.ai, the AI-native testing & observability platform used by teams shipping voice agents to handle thousands of real calls per day.

Special guests TBA — speakers from LiveKit, ElevenLabs, and/or Deepgram (confirming this week).

Pizza, drinks, and post-workshop networking included. Limited to 50 builders — RSVP early.

This event is hosted at the Frontier Tower:

We are transforming a 16-floor tower in San Francisco into a self-governed vertical village—a hub for frontier technologies and creative arts. Tier-one labs presenting AI, Ethereum, biotech, neuroscience, longevity, robotics, makerspace, human flourishing, and arts & music. These floors will house innovators and creators pushing the boundaries of human potential in a post-AI-singularity world.

Apply here for founding citizenship: https://frontiertower.io/apply

Why should I become a citizen?

  • Be part of creating the first self-governed vertical village

  • Connect with the most creative people in the city

  • Get access to all floors, free event space & movement floor

Website: https://frontiertower.io/

Need more reading? Visit https://frontiertower.notion.site/

Location
Frontier Tower @ 10th Floor Annex 995 Market Street, San Francisco
Avatar for Frontier Tower SF
Presented by
Frontier Tower SF
A 16-floor nexus for frontier tech in SF. Join us at https://frontiertower.io/apply
3 Going