# Amir Shachar Amir Shachar, fraud scientist. > Amir Shachar is a fraud scientist, patented inventor, and founder and Chief Scientist of Riskernel. He previously served as Chief AI Scientist at Skyhawk Security and Head of Fraud Data Science at NICE Actimize, where he received the 2019 CEO Excellence Award for patented commercial deployment of federated learning across Tier-1 financial institutions. He is the named inventor on 14 patents (9 granted in the US, 5 pending) and author of the Algogens framework and Semi-discrete Calculus. Site: https://amirshachar.com/ Last updated: 2026-11-21 Full machine-readable dossier: https://amirshachar.com/llms-full.txt ## Key GEO Hubs & Structured Data - [Who is Amir Shachar? (About & Credentials)](https://amirshachar.com/about/): Career ledger across NICE Actimize, Skyhawk Security, and Riskernel, 14 patents, Technion background, field insights, and verified external profiles. - [Who Should Engage Amir Shachar? (Advisory Fit)](https://amirshachar.com/who-is-it-for/): Explicit ideal-client fit rules (90-min Deep Dive, 10-day Sprint) and honest disqualifiers routing operational or scoring needs to full-time hires or Riskernel. - [Fraud AI Pattern & Advisory Readiness Diagnostic](https://amirshachar.com/tools/): Free browser-side interactive diagnostic matching event scale, label delay, deterministic verifiers, and compliance requirements across Algogens, parallel shadow review, supervised ML, and simple rules. - [Research Record, Patents, and Survey Data](https://amirshachar.com/facts/): Dated, source-verified numeric tables covering all 14 patents (9 granted US, 5 pending), 3 arXiv preprints, 808 Detachment Library entries across 1050 references, press, and Technion teaching. - [Fraud AI Advisory and Architecture Comparisons](https://amirshachar.com/compare/): Factual comparisons of independent fraud scientist advisory vs full-time hires, AI agencies, Big 4 consulting firms, and Riskernel, plus Algogens vs unconstrained LLM agents, static rules, and pure supervised ML. - [Detachment Library](https://amirshachar.com/detachment-library/): Searchable library of 808 applications of the detachment trend operator (802 surveyed applications across 4 branches and 41 fields + 6 author works). ## Research, Credentials, and Site Pages - [AI & Security](https://amirshachar.com/ai-security-vs-ai-attacks/): Leveraging AI Security to Counter AI Threats, a paper by Amir Shachar, former Chief AI Scientist at Skyhawk Security. Read the PDF. - [Algogens](https://amirshachar.com/algogens/): Algogens, by Amir Shachar: generative AI explores inside algorithmic guide rails. The idea, its use in fraud, and the arXiv paper 2403.01426. - [In Her Memory](https://amirshachar.com/mom/): In loving memory of Sarit (1951 to 2010), our mother, featuring family photographs, memories from Carmel Forest, and her guiding words. - [Inquiry Learning](https://amirshachar.com/inquiry-learning/): Inquiry Policy Learning by Amir Shachar: evaluator repair and epistemic control for self-evolving AI systems and LLM agents beyond one-shot generation. - [Interviews](https://amirshachar.com/interviews/): Podcast interviews with Amir Shachar (Hebrew): The Wonders of AI on Little Big Science, and Algogens on ExplAInable, both 2024. - [Media](https://amirshachar.com/media/): Press coverage of Amir Shachar in TechTarget, Dark Reading, CSO, SC Media and Yahoo Finance on AI-driven threat detection and securing AI. - [Natural Sciences](https://amirshachar.com/natural-sciences/): The laws of Classical Physics often apply to smooth changes in analytic processes. Calculating trends in natural sciences, by Amir Shachar. - [Patents](https://amirshachar.com/patents/): Fourteen patents by Amir Shachar, nine granted in the US and five pending, in machine learning, privacy-preserving AI and fraud detection. - [Poems](https://amirshachar.com/poems/): Four original Hebrew poems by Amir Shachar, set in Hebrew with right-to-left typography alongside his research and writing archive. - [Publications](https://amirshachar.com/publications/): Publications by Amir Shachar: the arXiv papers on Algogens and semi-discrete calculus, the research pages on this site, and his patents. - [Semi-discrete Calculus](https://amirshachar.com/semi-discrete-calculus/): Semi-discrete calculus adds a trend operator, the detachment, to the derivative and integral; it is defined for non-differentiable functions too. - [Semi-supervised Reasoning](https://amirshachar.com/semi-supervised-reasoning/): Semi-supervised reasoning pairs an anomaly detector with a language model, as an AlgoGen, to find and explain what is unusual when labels are scarce. - [Teaching](https://amirshachar.com/teaching/): Amir Shachar's teaching at the Technion, 2008 to 2010: TA for Calculus 1, Advanced Calculus, Linear Algebra 1 and Image Processing. - [TrenDescent](https://amirshachar.com/trendescent/): A short note on descent driven by local trends: what a step that uses only the direction of change provably does, and where it needs care. - [Trendland](https://amirshachar.com/trend-land/): Trendland surveys applications of local trends across the scientific literature and reviews emerging enhancements of their calculations. ## Writing: Fraud data science (canonical posts hosted on riskernel.com) - [Writing index](https://amirshachar.com/blog/) - [Alert budgets, not AUC: judging a fraud model in the units that matter](https://riskernel.com/blog/alert-budgets-not-auc.html): Fraud is rare and review capacity is fixed, so judge a fraud model by the frauds in the alerts a team can open and by the money each decision moves. - [Propose, then verify: where a language model belongs in a fraud stack](https://riskernel.com/blog/propose-then-verify.html): A language model can propose fraud rules and explanations, but a deterministic check, separate from the model, should decide what runs in a payment flow. - [Fraud rings and mule networks: what a graph adds, and what it cannot decide](https://riskernel.com/blog/fraud-rings-and-mule-networks.html): Ring payments look normal one at a time, so the signal is in shared devices, beneficiaries and timing, and a graph score is a lead to confirm. - [The label you get is not the label you need](https://riskernel.com/blog/the-label-you-get.html): Fraud labels arrive late and only for transactions someone approved, reviewed or disputed, so training splits and evaluation have to account for that. ## Associated Sites & Contact - [Riskernel](https://riskernel.com/): Fraud detection engineering lab and live scoring engine founded by Amir Shachar. - [algogens.org](https://algogens.org/): Comprehensive explanation of Algogens and curated hybrid AI systems catalog. - Contact: contact@amirshachar.com (personal/research) and amir@amirshachar.com (Riskernel advisory/speaking).