Four open-source AI agents for financial services — click through and run them against pre-loaded sample data. Each trial runs the real agent against Claude, capped at a few runs per session so it stays free to host.
Multi-agent AML alert triage system — enrichment, risk scoring, SAR drafting, HITL queue, and audit trail for financial crime compliance. Compliance analyst turnover is a significant hidden cost. AML Copilot removes the monotonous parts of the job — repetitive false-positive dismissals, manual data gathering, SAR drafting from scratch — leaving analysts to do the intellectually engaging work they were hired for. Job satisfaction improves. Institutional knowledge is retained.
Trial: pre-loaded sample alerts only · 3 runs per session
KYC compliance is one of the most expensive, slowest, and most error-prone processes in financial services. Analysts spend 80% of their time aggregating data, not assessing risk. This tool automates all of it — sanctions screening, PEP checks, adverse media, company registry lookups, and country risk — in under 60 seconds per entity.
Trial: pre-loaded sample entities only · 3 runs per session
Multi-agent AI model risk management platform — reads a Model Development Document and produces SR 11-7 compliance gap analysis, validation questions, risk rating, and a report draft. In under 2 minutes. Model Risk Management (MRM) is a regulatory compliance practice in banking and finance used to manage, minimize, and govern the financial or reputational losses caused by flawed computer models.
Trial: pre-loaded sample MDDs only · 3 runs per session
An AI-powered, multi-agent data governance platform for global banking institutions. Three specialised Claude agents — Metadata, Lineage, and Data Quality — work together to automate the most time-consuming parts of data governance: cataloguing datasets, mapping field-level lineage, and profiling data quality.
Trial: pre-loaded sample datasets only · Haiku model · 3 runs per session
Pick a sample case — no upload, no manual entry. Every trial runs on pre-loaded data so nothing you type ever reaches the model.
The agent runs for real against Claude — the same pipeline as the full repo, just capped to a cheap model and a few runs per browser session.
Want to run your own data? Clone the repo — every trial links straight to its GitHub source.
I'm open to Senior / Lead Product Manager roles in financial services, fintech and regulated industries — particularly where data, AI and platform modernisation meet.