FinanceNova.ai

The problem shaped the architecture.

Private banks want AI-assisted advisory, but every recommendation must be explainable, suitable for the client and available for audit. Generic copilots produce answers, not the evidence chain that compliance teams need.

An advisor needs a concise recommendation during a client conversation, while compliance needs the slower and more exact record behind it: which context was used, which rule applied, what the model generated and who reviewed it. FinanceNova was shaped around both users so that regulatory evidence is produced as part of the advisory workflow instead of reconstructed after the fact.

From product decision to working system.

I shaped the product architecture and AI agent layer, built the Vue frontend and mapped model output to MiFID II suitability requirements. I also designed the traceability workflow and the demo pipeline used in early fundraising.

The workflow captures the advisor request and client context, generates a recommendation, attaches provenance and suitability evidence, and routes higher-risk outputs through human review. Each transition contributes to an append-only audit record that can be inspected or exported without asking the model to explain itself retrospectively.

The decisions that hold the system together.

1. 01

   Evidence alongside generation

   Sources, model context and suitability checks are stored with the recommendation that used them.

2. 02

   Human review as a state

   Escalation and approval belong to the workflow model rather than an informal external process.

3. 03

   Reconstructable records

   The audit trail captures enough context to explain the path from advisor query to delivered recommendation.

From architecture to a verifiable result.

The work produced a pre-seed product narrative, a demonstrable advisory workflow and a compliance model that connects each recommendation to its sources, review state and regulatory evidence.

The useful part is what the work taught me.

In regulated products the audit trail is part of the user experience. Advisors need useful guidance while compliance teams need provenance, review states and exports that survive scrutiny.