Problem Statement
Delivering personalised investment guidance manually is resource-intensive and creates compliance risk. Figg Wealth needed to automate advice, risk profiling, and document handling at scale while maintaining strict alignment with MiFID II and FCA standards, producing consistent, auditable outcomes, with appropriate guardrails on what an AI is allowed to say.
Proposed Solution & Architecture
We designed Ask Figg as a cloud-native platform on Amazon EKS, orchestrating an Amazon Bedrock agent for conversational wealth guidance alongside a deterministic risk-profiling engine. The platform combines large-language-model reasoning with rule-based regulatory logic, a polyglot data layer for structured and document data, and defence-in-depth security throughout.
Ask Figg Conversational Agent
An Amazon Bedrock agent (Anthropic Claude) delivers conversational wealth guidance grounded in the user's profile and Figg's knowledge base, with guardrails constraining scope, tone, and the boundaries of regulated advice.
Automated Risk-Profiling Engine
A deterministic scoring engine calculates Risk Capacity and Risk Appetite from a structured MiFID II / FCA-aligned questionnaire, classifying users into five regulatory risk categories with consistent, auditable results.
Document Intelligence & Valuation
Amazon Nova Pro and supporting services process uploaded documents and asset-valuation inputs, with sensitive files served only through secure, time-limited presigned URLs, never public exposure.
Polyglot Data Layer
Amazon Aurora Serverless v2 (PostgreSQL), Amazon DocumentDB, and Amazon DynamoDB each hold the data they are best suited to: relational records, documents, and high-throughput key-value state respectively.
Security & Compliance
JWT authentication enforced at the API and application layer (no Amazon Cognito), AWS WAF and Amazon CloudFront at the edge, AWS KMS encryption, and least-privilege IAM, with an audit trail supporting regulatory readiness.
Scalable by Design
Containerised workloads on Amazon EKS scale with demand, while managed and serverless data services absorb growth without re-architecture.
AWS Services & Technologies
What We Delivered
- Designed and built the Ask Figg platform on Amazon EKS, orchestrating an Amazon Bedrock agent for conversational wealth guidance with scope and safety guardrails.
- Built the automated risk-profiling engine with Figg's compliance stakeholders, mapping MiFID II requirements to questionnaire inputs and scoring weights, calculating Risk Capacity and Risk Appetite independently.
- Implemented document intelligence and asset-valuation handling using Amazon Nova Pro, with sensitive files served only through secure, time-limited presigned URLs.
- Designed a polyglot persistence layer: Amazon Aurora Serverless v2 (PostgreSQL), Amazon DocumentDB, and Amazon DynamoDB, matched to each data type.
- Implemented defence-in-depth security: JWT authentication at the API and application layer (no Amazon Cognito), AWS WAF, Amazon CloudFront, AWS KMS encryption, and least-privilege IAM with audit logging.
- Validated scoring consistency and advisory guardrails across user scenarios for reliable, repeatable outcomes.
Outcomes & Success Metrics
- Personalised wealth guidance delivered conversationally at scale, grounded in each user's profile, with guardrails on the boundaries of regulated advice.
- Investor risk profiling automated: consistent, auditable Risk Capacity and Risk Appetite outcomes mapped to five FCA-aligned categories.
- Document and asset-valuation handling automated with secure, time-limited access: no public exposure of sensitive files.
- Polyglot data layer and Amazon EKS workloads scale with user growth without re-architecture.
- MiFID II and FCA-aligned architecture and outputs reviewed with compliance stakeholders.