HN Showcase
Back to browse
Veil – On-device meeting transcription with 100% PII redaction

Veil – On-device meeting transcription with 100% PII redaction

by nakaiwilliams·Mar 6, 2026·2 points·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerZero to OneBig BrainSolve My Problem

Custom PII model runs on-device before AI sees anything—solves real data exposure in meeting tools.

Strengths
  • •Genuinely novel architecture: local PII redaction before cloud processing eliminates the privacy problem that other AI meeting tools accept.
  • •Custom 72M-word PII model with claimed 100% catch rate demonstrates serious training effort and domain specificity.
  • •Solves real compliance pain (legal privilege, medical confidentiality) where existing tools (Otter, Fireflies) fail by design.
Weaknesses
  • •macOS notarization ($99) not yet completed; Windows support exists but may lag in polish.
  • •Solo dev early-stage project; no SLA, long-term support, or enterprise contracts visible.
Category
Productivity
Target Audience

Legal firms, medical practices, executives handling sensitive conversations, enterprise compliance teams

Similar To

Otter.ai · Fireflies.ai · Grain

Post Description

Solo dev from South Africa. Spent 3 months building this.

Every AI meeting tool sends conversations to the cloud. I trained my own PII model - 72M words, 100% catch rate on real meetings. Before anything touches AI, names become aliases, amounts scramble, identifiers vanish.

Everything runs on-device. Nothing leaves your machine.

Free for now - can't afford Apple's $99 notarization yet. Happy to answer questions about the tech.

Similar Projects