Major project
Banking & Fintech
Secure banking assistant in production
Deploy a customer assistant with personal-data masking, refusal of risky requests and a security review.
The brief
A bank wants an AI assistant for customers but its risk team fears data leaks and manipulation. Build and deploy an assistant that answers banking questions, masks personal data, refuses risky requests, and passes a security review against the OWASP Top 10 for LLM applications. Good looks like: a deployed assistant, a red-team report with fixes, usage and cost dashboards, and a demo for a panel.
What you will build
- Define what the assistant may and may not do
- Build it with grounding in FAQs and product terms
- Mask personal data in inputs, logs and outputs
- Red-team it against the OWASP Top 10 for LLM applications and fix the findings
- Deploy with usage, quality and cost dashboards
- Present the risk review and results to a panel
Free, open data to build it with
- Banking77 customer queries ↗ Hugging Face (PolyAI)
- OWASP Top 10 for LLM applications ↗ OWASP
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.