# Vantiq Studio

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Aun Abidi](https://cerebralvalley.ai/u/aunabidi)
- **GitHub:** https://github.com/aunabidi-dev/VelatoAI
- **Demo video:** Https://drive.google.com/file/d/1ZQ3dojBXXBJJK6nyRXseW-AoTa500t45/view?usp=drivesdk
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3713

Velato AI: Zero-Trust Voice Payment Capture and Live Support Copilot

When a customer pays by phone, an AI takes over so no human ever hears or sees the card.

The problem
Contact-center agents hear customers’ card numbers all day. Even with strict floor rules, an agent can memorize a number and misuse it later. Keypad (DTMF) masking is clumsy and can break on VoIP and WebRTC calls. Call recordings, kept for quality assurance, can leak the same numbers to reviewers. The idea came from a friend who worked at a large BPO and saw this risk firsthand.

What it does
Velato AI is a module that sits on the call. It is not a CRM.

Live copilot: Muse Spark Open turns the conversation into text and, using documents and scripts uploaded by an admin, shows suggested replies, policy references and next steps on the agent’s screen. New agents get productive faster.
PCI Safe Mode: the agent tells the customer they are being passed to a secure automated system and presses one button. The server cuts the agent’s audio path, the copilot goes blind and recording pauses. An AI voice agent introduces itself, collects the card number, expiry and CVV by voice, checks the number with the Luhn algorithm, reads back only the last four digits and the expiry, and obtains a token. The agent returns to the call seeing only the token and the last four digits.
Recordings stay safe: one continuous recording per call, but during payment it writes a short notice instead of live audio, so QA reviewers cannot hear the card either.
Nothing card-related is stored. Card data lives in memory for seconds and is wiped. An automated scanner checks the database, logs and transcripts for card-like numbers.

## More from Meta Global AI Developer Hackathon

- [ChronoRig](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3710)
- [Wazi](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3711)
- [metalmind](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3712)
- [WorkLens](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3714)
- [Machine Vision Radar](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3715)
- [Quinn Avila](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3716)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3713. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt