I work at the intersection of fraud investigation, AML compliance, and applied AI automation inside regulated financial environments.
My focus is on building LLM-assisted workflows that reduce manual analytic load, improve consistency, and increase defensibility of decisions.
aml-sar-drafting-agent-suite— AI-assisted orchestration for consistent, policy-aligned SAR narrative draftingfraud-case-triage-orchestrator— Agentic pre-classification & enrichment layer for fraud case workflows before analyst reviewregflow-compliance-monitor— Continuous rule-conformance monitoring against internal controls & regulatory requirements
These projects reflect production-grade thinking: alignment to banking constraints, auditability, explainability, and risk posture.
AI systems that assist — not replace — regulated decision-making by:
- Pre-structuring fraud/AML inputs before analyst review
- Orchestrating multi-agent analysis against legal & bank rules
- Generating draft artefacts (e.g. SAR narratives) with audit-trace
- Embedding regulatory guardrails into automated flows
Architecture & Methods (click to expand)
- Agentic orchestration: CrewAI, strategy trees, context-scoped tool access
- Data stack: Python, SQL, ETL for investigation-grade inputs
- Reasoning constraints: policy-aligned prompt control, defensive LLM usage
- Workload targets: case triage, enrichment, narrative generation, rule checks
- Non-numerical success criteria: consistency, defensibility, throughput, auditability
Fraud / AML / Compliance Domain Scope (click to expand)
- Fraud case investigation & pre-arbitration workflows
- AML fundamentals: SAR drafting logic, red flags, typologies
- Governance & controls: internal rule adherence, explainability
- Regulatory context awareness (EU banking & financial crime)
Other Technical Areas (crypto analytics / market structure) — reduced scope
Applied crypto/market microstructure projects retained only where analytically transferable (e.g. liquidation heatmapping, automated capture logic, risk-driven alerting), now positioned as methodology, not career identity.
Open to roles in Fraud AI · Financial Crime Automation · Applied LLM Systems in regulated environments.
Email: koljakiss@icloud.com