AI-Driven Agent Framework for Quantum Financial Modeling and Optimization
Project Description
The intersection of Quantum Computing and Finance (Quantum Finance) offers a paradigm shift for solving complex, computationally intensive financial problems—such as portfolio optimization, derivative pricing, and algorithmic fraud detection. However, mapping intricate, real-world financial constraints into quantum algorithms (like QAOA or Quantum Neural Networks) requires deep, cross-disciplinary expertise.
This project aims to develop an intelligent, AI-driven assistant framework that utilizes Large Language Models (LLMs) and advanced AI agents to automate the formulation, translation, and execution of quantum financial models.
Key Objectives & Methodology:
> ntelligent Problem Formulation: Build LLM agents capable of ingesting raw, natural-language financial objectives (e.g., maximizing the Sharpe ratio under sector-exposure constraints) and automatically translating them into mathematically sound Quadratic Unconstrained Binary Optimization (QUBO) problems.
> Hybrid Orchestration: Implement an agentic workflow that connects these formulated models to quantum simulation backends (such as Qiskit or PennyLane) or quantum-inspired hardware, automatically tuning hyperparameters for optimal convergence.
> Empirical Benchmarking: Evaluate the AI-generated quantum workflows against industry-standard classical alternatives (e.g., traditional Markowitz Mean-Variance optimization, Monte Carlo simulations) using historical market data to analyze speedup, accuracy, and noise resilience.
This project aims to develop an intelligent, AI-driven assistant framework that utilizes Large Language Models (LLMs) and advanced AI agents to automate the formulation, translation, and execution of quantum financial models.
Key Objectives & Methodology:
> ntelligent Problem Formulation: Build LLM agents capable of ingesting raw, natural-language financial objectives (e.g., maximizing the Sharpe ratio under sector-exposure constraints) and automatically translating them into mathematically sound Quadratic Unconstrained Binary Optimization (QUBO) problems.
> Hybrid Orchestration: Implement an agentic workflow that connects these formulated models to quantum simulation backends (such as Qiskit or PennyLane) or quantum-inspired hardware, automatically tuning hyperparameters for optimal convergence.
> Empirical Benchmarking: Evaluate the AI-generated quantum workflows against industry-standard classical alternatives (e.g., traditional Markowitz Mean-Variance optimization, Monte Carlo simulations) using historical market data to analyze speedup, accuracy, and noise resilience.
Supervisor
FUNG, May
Quota
1
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
The selected applicant will act as a Research and Development (R&D) Engineer on this cross-disciplinary project. Specific responsibilities include:
> Agentic Framework Development: Writing Python-based workflows using LLM APIs and agent frameworks to parse financial constraints and output mathematical optimization models.
> Quantum Backend Integration: Assisting in connecting the AI-generated formulations to quantum simulation environments (e.g., Qiskit, PennyLane) to run optimization algorithms.
> Data Engineering & Benchmarking: Collecting historical market data, running baseline classical financial simulations (Markowitz/Monte Carlo), and comparing their performance metrics against the quantum-hybrid outputs.
> Documentation & Reporting: Maintaining clean code repositories and documenting empirical results regarding algorithm accuracy and execution speedups.
> Agentic Framework Development: Writing Python-based workflows using LLM APIs and agent frameworks to parse financial constraints and output mathematical optimization models.
> Quantum Backend Integration: Assisting in connecting the AI-generated formulations to quantum simulation environments (e.g., Qiskit, PennyLane) to run optimization algorithms.
> Data Engineering & Benchmarking: Collecting historical market data, running baseline classical financial simulations (Markowitz/Monte Carlo), and comparing their performance metrics against the quantum-hybrid outputs.
> Documentation & Reporting: Maintaining clean code repositories and documenting empirical results regarding algorithm accuracy and execution speedups.
Applicant's Learning Objectives
Through this project, the applicant will achieve the following highly valuable competencies:
> Advanced AI & LLM Orchestration: Gain hands-on experience in building specialized AI agents, prompt engineering for mathematical formulation, and managing multi-agent workflows.
> Financial Engineering Fundamentals: Deepen understanding of quantitative finance concepts, including risk-return profiles, portfolio optimization metrics (Sharpe ratio), and classical simulation techniques.
> Introduction to Quantum Computing Applications: Learn how real-world optimization problems are translated into quantum-compatible formats (QUBO) and executed on quantum simulators using industry-standard SDKs like Qiskit.
> Empirical Research Methodology: Develop rigorous benchmarking skills by comparing emerging technologies against established classical baselines using real-world historical data.
> Advanced AI & LLM Orchestration: Gain hands-on experience in building specialized AI agents, prompt engineering for mathematical formulation, and managing multi-agent workflows.
> Financial Engineering Fundamentals: Deepen understanding of quantitative finance concepts, including risk-return profiles, portfolio optimization metrics (Sharpe ratio), and classical simulation techniques.
> Introduction to Quantum Computing Applications: Learn how real-world optimization problems are translated into quantum-compatible formats (QUBO) and executed on quantum simulators using industry-standard SDKs like Qiskit.
> Empirical Research Methodology: Develop rigorous benchmarking skills by comparing emerging technologies against established classical baselines using real-world historical data.
Complexity of the project
Challenging