Machine Learning |

AI Application in FinTech

Fri, 17 Jul 2026 00:00:00 +0000

课程简介

本课程面向具有基础数理背景的学生，系统介绍人工智能在金融科技领域的技术原理与实际应用。课程分为两大部分：第一部分从图灵测试出发，梳理 AI 技术的演进脉络，涵盖监督学习、无监督学习、强化学习三大范式，深入讲解神经网络、注意力机制、Transformer 架构、大语言模型（LLM）以及检索增强生成（RAG）和 AI Agent 等前沿概念。第二部分聚焦金融科技应用场景，包括智能投顾（Robo-Advisory）、组合优化与现代投资组合理论、直接指数化（Direct Indexing）、税收亏损收割（Tax-Loss Harvesting），以及如何将多角色 AI Agent 系统应用于投资研究与决策流程。课程通过真实案例和实操演示，帮助学生理解如何将机器学习与大模型能力转化为可部署的金融产品与服务。

资料下载

AI-Driven Automatic Wealth Management Solution

Sun, 31 Mar 2024 00:00:00 +0000

Period: April 2021 – March 2024
Funding: Innovation and Technology Fund (ITF), Hong Kong, Reference No. B/E003/20
Industry partner: Magnum Research Limited (AQUMON)

Problem

End-to-end digital wealth advisory requires integrating client profiling, goal-based portfolio construction, ongoing rebalancing, and monitoring within a scalable, auditable, and compliance-aware system. Existing approaches either rely on static rule-based logic or lack the modularity required for institutional deployment across diverse client segments and regulatory jurisdictions.

Method

The project developed a full-stack AI advisory architecture covering: (i) client profiling and suitability assessment using ML classifiers; (ii) goal-based portfolio construction integrating intertemporal allocation theory with modern estimation techniques; (iii) a hybrid-cloud data foundation supporting global asset universes and real-time feeds; (iv) high-fidelity backtesting with controls against overfitting; and (v) integration with execution infrastructure. Research components addressed model selection, uncertainty quantification in portfolio optimisation, and the design of modular research-validation-deployment pipelines.

Outcome

The project delivered a production-grade automated wealth management architecture deployed across multiple institutional client implementations. It established a reusable technical foundation for AQUMON’s robo-advisory and AI advisory services.

ESG Quantitative Scoring System and ML-Driven Investment Research

Sun, 01 Jan 2023 00:00:00 +0000

Problem

Existing ESG ratings from major providers suffer from low cross-provider agreement, limited transparency, and methodological subjectivity. A more systematic, data-driven approach is needed that integrates structured financial data with unstructured news and disclosure content, updates at high frequency, and supports both index construction and active strategy development.

Method

AQUMON developed a four-layer evaluation framework — Pillar, Category, Topic, and Metric — covering environmental, social, and governance dimensions for a universe of listed companies. The system integrates structured financial and regulatory data with unstructured news, corporate disclosures, and sentiment signals using multimodal ML pipelines. Dynamic weight updating and cross-industry comparability were explicit design requirements. The scoring methodology was designed to satisfy IFRS S1 and IFRS S2 disclosure frameworks and supports both institutional investment mandates and corporate sustainability reporting.

Outcome

The system provides daily ESG score updates covering 3,000+ listed companies. Backtested long-only strategies combining ESG scores with fundamental signals demonstrated improved risk-adjusted performance relative to market benchmarks. The framework has been adopted by the HKSAR Government and multiple listed companies for sustainability reporting and investment mandate compliance.

Portfolio Management via Reinforcement Learning

Fri, 31 Dec 2021 00:00:00 +0000

Period: January 2019 – December 2021
Funding: Innovation and Technology Fund (ITF), Hong Kong, Reference No. UIM/365
Industry partner: Magnum Research Limited (AQUMON)

Problem

Standard portfolio rebalancing rules are based on fixed-weight targets and threshold-trigger mechanisms that do not adapt to evolving market conditions or explicitly account for transaction costs. A principled sequential decision-making framework is needed to learn rebalancing policies that remain effective under diverse market regimes and realistic cost structures.

Method

The portfolio rebalancing problem is formulated as a Markov Decision Process (MDP) with state variables capturing portfolio composition, market features, and risk indicators; actions corresponding to rebalancing decisions; and a reward function that penalises both tracking error and transaction costs. We explored value-based and policy-gradient reinforcement learning algorithms, including recurrent and sequence-based architectures that condition on the trajectory of market states. Emphasis was placed on sample efficiency, robustness to distribution shift, and deployability in a production rebalancing pipeline.

Outcome

The project produced a research-to-production foundation for RL-based portfolio management, with documented methodology, experimental benchmarks, and a reusable framework for systematic strategy development and evaluation at AQUMON.