Executive Summary

Bunq Investment Coach is an AI-powered investment advisory platform that demystifies investment decisions for everyday banking customers. Developed during a hackathon in partnership with Bunq, this solution analyzes users' transaction histories to create personalized investment profiles, portfolio recommendations and rebalancer. The platform features an intuitive chat interface that allows users to understand and adjust their investment strategies in natural language, making investing accessible to everyone regardless of financial literacy level.

Problem Statement

Investing remains intimidating and complex for average consumers:

Many individuals lack confidence in selecting appropriate investments Traditional advisory services are expensive and inaccessible Most people don't understand how their spending patterns relate to investment potential Investment terminology and concepts create significant barriers to entry Timing investment decisions feels like guesswork for novice investors

Our Solution

Bunq Portfolio Rebalancer bridges the gap between banking and investing by:

Analyzing existing financial behavior through Bunq transaction data Generating personalized investment profiles based on spending patterns Recommending optimized portfolios aligned with user risk tolerance Providing an AI chat interface for investment guidance and adjustments Visualizing potential returns to make abstract concepts tangible

Architecture Specification

Technology Stack

Backend: Elixir/Phoenix
Database: PostgreSQL
Authentication: API Key-based (X-API-Key header)
LLM Integration: Using LLMs for risk profiling and portfolio suggestions
Additional Tools: Tailwind CSS, Sobelow, Credo, Ecto, ExUnit, Plug, Phoenix LiveView

System Components

1. API Service

RESTful API implementation following OpenAPI 3.0.3 specification
Endpoints structured around profile → portfolio → chat workflow
JSON-based request/response formats
Error handling with standardized error codes
Implementation using Phoenix framework

2. Transaction Processing

Simple transaction data ingestion from Bunq API
Basic data cleaning and categorization
Ecto schemas for data persistence
Modular design for future extensibility

3. LLM-based Analysis

Risk Profile Agent: Analyzes transaction data to determine risk profile and score
Portfolio Suggestion Agent: Uses risk profile, balance, and available assets to suggest a portfolio
No complex ML models or algorithms
Simple prompt templates for consistent outputs

4. Chat Interface Backend

Simple context-aware response generation
Portfolio and profile update capabilities
Stateful conversations through database persistence
Integration with frontend components

Component Interaction

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│                 │    │                 │    │                 │
│  Bunq API       │───>│  Transaction    │───>│  Risk Profile   │
│                 │    │  Processing     │    │  Agent (LLM)    │
└─────────────────┘    └─────────────────┘    └────────┬────────┘
                                                       │
                                                       ▼
┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│                 │    │                 │    │                 │
│  Frontend UI    │<───│  API Service    │<───│  Portfolio      │
│                 │    │                 │    │  Agent (LLM)    │
└─────────────────┘    └─────────────────┘    └─────────────────┘
        ▲                      ▲
        │                      │
        └──────────────────────┘
             Chat Interface