This Startup Catches Fraud at Scale

Variance co-founder Karine Mellata announces $21M Series A and reveals how a 12-person team automates fraud detection for Fortune 500s and GoFundMe using AI agents.

• Variance raised $21M Series A after 3 years in stealth; clients include GoFundMe and unnamed Fortune 50 companies.
• During elections, Variance AI agents detected state-sponsored coordinated fraud rings — something isolated classifiers could not have caught.
• The core technical blocker for years was unstructured data scattered across 5–10 internal systems, sometimes only accessible by scraping human-facing UIs.
• AI agents replaced rules engines, classifiers, and human analysts entirely; the system is self-healing with no slow feedback loop.
• 5 engineers output at the level of ~25 by running coding agents on three monitors each; a non-technical CSM ships features via Cursor without touching the eng team.
• First customer IAC (care.com / Ask Media Group) took 8 months to land; GPT-4 dropped mid-pilot and cut costs 10x while changing performance.
• In July 2024, Karine broke her spine and leg after being hit by a truck during peak growth (revenue doubling month-over-month); the company survived on Michael running engineering while she recovered.

2026-03-31 · Watch on YouTube