I build and evaluate generative-AI systems for regulated financial services - using NLP and LLMs to improve client-documentation quality, KYC, identity matching, and fraud detection, with risk management and compliance built in from the start.
Approach
Methodology
1
Frame the regulated banking use case
Start from a concrete compliance problem - client-documentation quality, KYC, identity matching, or fraud detection - and pin down the risk and regulatory constraints it has to satisfy.
2
Build the NLP / LLM solution
Apply text analytics and LLMs (Anthropic/Claude and OpenAI APIs) to assess and improve documentation, calling them against an explicit quality and compliance rubric.
3
Embed risk and compliance controls
Bake risk-management and compliance checks into the workflow so generated and assessed content stays auditable and defensible, not just fluent.
4
Validate and communicate results
Evaluate solutions to a research standard and communicate findings to stakeholders - from hackathon judges to peer-reviewed publication.
In Practice
Code Snippets
note_qa.pypython
def assess_contact_note(note, llm, rubric):
"""Quality-assure a client contact note against a compliance rubric."""
prompt = build_prompt(note=note, rubric=rubric)
review = llm.complete(prompt) # Anthropic/Claude or OpenAI
return {
"completeness": review.completeness,
"compliance_flags": review.flags,
"suggested_edits": review.edits,
}
quality_rubric.txttext
Assess this client contact note for a Swiss private bank.
Check completeness, factual support, and regulatory compliance.
Flag missing KYC details, unsupported claims, and risk indicators.
Return JSON: {"completeness": 1-5, "flags": [...], "edits": [...]}.
Results
Case Studies
RiskOn Hackathon 2025 - Julius Baer
Problem
Client contact notes vary in quality, creating compliance and risk exposure for the bank: 'Quality Assurance for Client Contact Notes - how can AI help?'
Approach
Built an NLP/LLM solution that applies text analytics to automatically quality-assure client contact notes, surfacing gaps and compliance risks for review.
Result
Team Winner of the RiskOn Hackathon 2025, sponsored by Julius Baer (Bank), Switzerland.
AI-driven risk management in Swiss banking
Problem
Swiss banks need stronger, AI-enabled approaches to client documentation, identity matching, and fraud detection across the KYC lifecycle.
Approach
Researched and designed AI-enabled solutions for KYC, identity matching, and fraud detection, grounding them in real risk-management and compliance requirements.
Result
Published as 'Advancing Risk Management in Swiss Banking' with the University of Zürich, Department of Finance (2026).