HOME  >  Research    
Wealth Playbook (WPB - AI Financial Advisor)

Introduction

The Wealth Playbook platform is an innovative investment planner designed to bridge the gap between individual investors and personalized financial advice. In the current landscape, many investors, particularly younger generations, seek portfolios that align with their personal beliefs, such as ESG (Environmental, Social, and Governance) principles or specific technology themes.

However, existing solutions in markets like Thailand are often "one-size-fits-all" or restricted to high-net-worth individuals. Wealth Playbook addresses this by providing a low-cost, web-based solution that converts complex user goals, risk profiles, and qualitative preferences into concrete, diversified, and explainable Exchange-Traded Fund (ETF) portfolios.

Key Features

  • Natural Language Interpretation: Unlike traditional robo-advisors that rely solely on multiple-choice questions, the platform uses Large Language Models (LLMs) to interpret free-text descriptions of goals and constraints.
  • Hybrid Asset Selection: The system combines rule-based filtering (to ensure risk and regulatory compliance) with semantic ranking (to match user intent with investment themes).
  • Explainable Investing: Every suggested portfolio comes with a plain-language explanation, detailing exactly why each asset was chosen to foster user trust and transparency.
  • Thematic & ESG Alignment: Investors can specify niche interests, such as "clean energy focus with AI exposure," and receive a portfolio tailored to those specific convictions.

Development and Innovation

The core of the Wealth Playbook is its Asset Selection Engine, which utilizes a sophisticated three-step pipeline:

  1. Interpretation Layer: Powered by Gemini 2.5 Flash, this layer extracts "investment kinds" from unstructured user input, identifying themes and risk levels.
  2. Semantic Search & Ranking: Using the all-mpnet-base-v2 model and Sentence-Transformers (SBERT), the engine converts text into numerical vectors. It uses Cosine Similarity to mathematically measure how well an ETF's description matches the investor's intent.
  3. Refinement & Review: A final LLM pass ensures the shortlisted ETFs respect diversification rules and constraints before passing them to a weight-optimization module.

Impact and Future Directions

Wealth Playbook aims to democratize personalized finance by providing high-tier advisory logic at a "robo-advisor" cost. By focusing the current development on a robust asset selection engine, the project lays the technical foundation for a full end-to-end platform.

Future goals include:

  • Expanding the curated ETF database to include global markets.
  • Integrating an advanced weight-optimization module to further refine portfolio performance.
  • Scaling the engine to support a broad base of Thai investors, enabling them to build wealth in a way that is both financially sound and personally meaningful.

Project Advisor(s)

Charnon Pattiyanon
Assistant Director of IT

Research Team member(s)

Natdanai Voraratanavivich
Undergraduate Student
Thanat Vithyanarakul
Undergraduate Student
Radit Srisathaporn
Undergraduate Student
Jesnaronk Jesadanont
Undergraduate Student