Essay

Why fraud models should only propose. By Amir Shachar.

Read the essay
Skip to content
Riskernel

Speaking and podcasts

Book Amir for podcasts, conference talks, and operator briefings on fraud infrastructure.

The strongest sessions are not generic fraud trend decks. They help an audience understand where fraud systems break in production, how serious teams should evaluate vendors, and why payout risk, setup risk, latency tails, and explainability for ops matter more than polished demos.

  • Featured in TechTarget
  • Featured in Dark Reading
  • Featured in CSO
  • Featured in Yahoo Finance

Talks

Talks hosts and organizers can actually use

Why fraud buying decisions fail

Most teams compare feature lists and vendor claims. The useful evaluation starts with shadow testing, threshold behavior, analyst usability, and what the stack does under real latency pressure.

What explainability actually means for ops

Per-decision reasons matter because analysts need something operationally usable, not a portfolio-level “AI transparency” story that collapses the moment a case hits the queue.

Why clean-looking payouts still become losses

The highest-cost fraud often hides in setup changes, first-time payees, payout timing, and context around release of funds, not just in the transaction event itself.

Best audiences

  • Fraud and risk leaders at fintechs, payment processors, digital banks, and BaaS platforms
  • CTOs and product leaders evaluating fraud infrastructure or replacing a legacy stack
  • Operators who care about approval friction, false positives, payout risk, and real-time decisioning

Why this lands with operators

  • Named inventor on 15 patent filings (9 granted US patents and 6 applications) in machine learning, privacy-preserving AI, and fraud detection; former Head of Fraud Data Science at NICE Actimize
  • Pioneered the commercial application and patented deployment of Federated Learning for Tier-1 fraud prevention at NICE Actimize (awarded the 2019 CEO Excellence Award)
  • Former Chief AI Scientist at Skyhawk Security
  • Architect of the AlgoGens framework (curator of the 100-system hybrid AI index) and author of the foundational research monograph on Semi-Discrete Modeling
  • Featured in TechTarget, Dark Reading, CSO, and Yahoo Finance
  • Current operator voice behind Riskernel’s fraud infrastructure essays and quarterly report

How to book

Send the audience, format, timing, and the outcome you want from the session. Podcasts and conferences should include the target audience, ideal angle, and expected time slot.