STAT GR5398 – MA Mentored Research (Spring 2026)

AI4Finance Open-Source Research Track

This repository hosts the official project hub for STAT GR5398 – MA Mentored Research (Spring 2026) at Columbia University, led by Bruce Yang, Founder & President of the AI4Finance Foundation.

This course is designed as a research-driven, open-source, and production-oriented mentored research experience. Students will work on real-world systems that are already used by researchers, practitioners, and industry teams in quantitative finance and AI.

All projects are based on large-scale open-source frameworks maintained by the AI4Finance Foundation and emphasize:

• Research depth
• Engineering rigor
• Real-world validation
• Open-source contribution

---

Course Structure

Students choose one of three research directions.
Each direction provides three project options, allowing flexibility while maintaining a unified research theme.

Important
This is not a tutorial-style course. Each direction requires sustained commitment (approximately 30 hours per week) and is suitable for students aiming for:

• Research publications
• PhD preparation
• Advanced industry roles in AI / quantitative finance

---

Direction 1: FinRL-Trading — Quantitative Trading Systems

Difficulty: High (Engineering-Intensive)

FinRL-Trading is a modern, modular quantitative trading platform built with Python, featuring machine learning and reinforcement learning strategies, professional backtesting, and live trading capabilities.

Open-Source Framework
https://github.com/AI4Finance-Foundation/FinRL-Trading

What You Will Do

• Design and implement end-to-end quantitative trading strategies
• Integrate ML / RL signals into the FinRL-Trading framework
• Run realistic paper trading (fake money) against live market data
• Evaluate performance, risk, and robustness under real market conditions

Expectations

• Strong Python engineering ability
• Solid understanding of financial markets
• Comfort working with complex, modular, production-level systems

Outcomes & Opportunities

• Outstanding contributors may receive:
  • Unpaid internship offers from AI4Finance-affiliated projects
  • Scholarship support for strategies demonstrating strong empirical performance
• Strong projects may evolve into open-source benchmarks or research papers

---

Direction 2: FinRobot — AI-Driven Equity Research

Difficulty: Medium

FinRobot is an open-source AI Agent platform for automated equity research and financial analysis using large language models (LLMs).

Framework
https://github.com/AI4Finance-Foundation/FinRobot

Live Demo
https://finrobot.ai

What You Will Do

• Build AI agents that generate professional-grade equity research reports
• Apply LLMs to:
  • Financial statement analysis
  • Business and industry modeling
  • Investment thesis generation
• Extend or customize the FinRobot agent architecture

Open-Source Commitment

• The entire FinRobot codebase will be fully open-sourced during the semester
• Students will build directly on production-level systems, not toy examples

Expected Output

• A complete, reproducible Equity Research Report
• Clear methodology, prompt design, and evaluation logic
• Optional research-style write-up suitable for publication

---

Direction 3: FinGPT — Financial Sentiment Analysis with LLMs

Difficulty: Very High (Data + Modeling + Finetune Focus)

FinGPT is an open-source financial large language model project focused on real-world financial NLP tasks.

Framework
https://github.com/AI4Finance-Foundation/FinGPT

This Semester’s Focus

We will build a financial data platform that integrates:

• News
• Social media
• Buy-side / institutional opinions

Using this data, students will:

• Design data pipelines and labeling strategies
• Fine-tune FinGPT models for financial sentiment analysis
• Validate model performance in real or near-real market settings

Skills Emphasized

• Data engineering and curation
• LLM fine-tuning and evaluation
• Financial NLP and sentiment modeling

Expected Output

• A cleaned and well-documented dataset (or subset)
• A fine-tuned FinGPT model
• Empirical evaluation linked to real market signals

---

Open-Source & Research Philosophy

All projects:

• Are open-source by default
• Follow research-grade standards (reproducibility, documentation, evaluation)
• Encourage upstream contributions to AI4Finance repositories

Students are expected to think and operate as:

• Researchers
• Engineers
• System builders

---

Evaluation & Mentorship

Evaluation is based on:

• Technical depth
• Research rigor
• Code quality
• Original contribution

Top-performing students may:

• Be invited to co-author research papers
• Receive long-term research mentorship
• Continue with AI4Finance projects beyond the semester

---

Who Should Take This Course

This course is ideal for students who:

• Seek real research experience, not simulated coursework
• Are comfortable working with large, evolving codebases
• Aim for PhD programs, research labs, or advanced industry roles
• Are willing to commit significant time and effort

---

Links

• Columbia MA Mentored Research (STAT GR5398):
  https://ma.stat.columbia.edu/stat-gr5398-ma-mentored-research/

• Course GitHub (Spring 2026):
  https://github.com/AI4Finance-Foundation/STAT-GR5398-Spring-2026

• AI4Finance Foundation:
  https://github.com/AI4Finance-Foundation