John L
- Research Program Mentor
MS at University of Michigan - Ann Arbor
Expertise
Artificial Intelligence, Machine Learning, Business, Finance, Economics, Business Analytics, Statistics, Psychometrics, Research Methods
Bio
I am a Statistics Lecturer at the University of Chicago. There, I designed a course on Data Science in Quantitative Finance and Risk Management and taught it to undergraduate students of all majors and backgrounds. What I enjoy most about teaching is that I am able to engage with students on both individual and team levels. My vision is to create a learning community for students to embrace uncertainty, obstruction, and frustration throughout their learning journey. I am also a Research Statistician at the College Board. My day-to-day tasks range from designing psychometric approaches to improving digital learning and assessments through a programmatic lens. I enjoy this field of work because I am able to communicate the technical facet of my work to field laymen. My goal is to novelize learning tools to help K-12 students make critical decisions in life.Project ideas
AI-Powered Quantitative Risk Dashboard App
Project Description: Ever wonder how Wall Street pros estimate how much money they could lose on a bad day, before it happens? You'll build an interactive web platform where users choose a risk personality, load a real or hypothetical portfolio, and go head-to-head against the machine to answer questions like "What's the chance this stock drops 10% next month?" Along the way you'll wield the same weapons professional risk managers use, such as Value-at-Risk, Expected Shortfall, Bayesian modeling, and even a neural network that can "look at" charts to turn scary market math into a game anyone can play. Project Outcomes: - Build the dashboard experience: portfolio entry, a risk-personality selector (risk-averse vs risk-taking), and results screens users actually enjoy using. - Implement the headline risk metrics: Value-at-Risk ("what's the most I'd expect to lose, 95% of the time?") and Expected Shortfall ("and when things DO go wrong, how bad is it on average?"). - Model the messy real world: fit heavy-tailed probability distributions (Student-t, Double Exponential, GED…) to market returns, and use Bayesian computing to estimate the distribution that best matches what's actually observed. - Arm users with decision tools: QQ plots, Kullback–Leibler (KL) divergence scores, scatter matrices, and correlation heatmaps, so they can pick their model like a pro, then compare their risk call against the machine's. - Crack portfolio optimization: use Mean-Variance Optimization and quadratic programming to find the mix of assets (long and short) that maximizes the Sharpe ratio or minimizes volatility. - AI implementation: teach a computer to read charts: train a CNN to evaluate QQ plots and efficient-frontier graphs (is this reward-to-risk tradeoff actually balanced?), including data augmentation to generate enough portfolio scenarios to train on. - User Interpretation: translate the machine's insights into friendly, personality-matched feedback that nudges users toward better distribution choices and smarter portfolio weights. - At the end of the program, the student will deliver a working web app prototype that qunatifies the uncertainties of a trading porfolio through probabilistic modeling and algorithmic optimization. Moreover, they will incorporate visualization techniques to make the technical implications easily understandable to any audience.
Agents of Alpha App Development
Project Description: What if you could hire an entire hedge fund team, except every "employee" is an AI agent you designed? You'll build an interactive web platform where users draft their dream team of AI agents, hand them an investing philosophy, and backtest the squad on historical market data to see whose fund would have crushed it. An example can be for them to train a "Warren Buffet" agent to select stocks, followed by a "Jim Simons" agent to trade them, then a "Ray Dalio" agent to manage the in-trade risks. Think fantasy football meets Wall Street, users can use your platform to simulate investing and trading strategies created by their agentic hedge fund team before deploying the team to make real financial decisions. Project Outcomes: - Build the dashboard experience: a "draft your team" screen, strategy configuration, a backtest mission control, and results dashboards. - Invent a roster of AI agent personas — value investor, technical chartist, risk manager, etc., each with its own prompt, personality, and decision-making style. - Write the backtesting engine: pull historical prices, simulate trades, track P&L, and score performance with metrics like total return, Sharpe ratio, and max drawdown. - Visualize the drama: equity curves, trade logs, agent "decision diaries," and team performance boards. - At the end of the program, the student will deliver a working web app prototype that trains and teaches the AI agents to behave like real-life traders.
AI-Powered Trading Journal Dashboard App
Project Description: Turn trading habits into data, and data into smarter decisions This project challenges students to build an intelligent trading journal that does more than just store trades - it thinks with you. Users can log trades with details like entry price, exit price, strategy used, and confidence level. The platform automatically calculates key stats such as holding time, profit/loss, and reward-to-risk ratio. Students will also design features that capture the human psychology side of trading, including emotions, stress levels, and decision confidence. The exciting twist? Students will train AI models to: 1) Detect patterns in a trader’s behavior 2) Spot risky habits or emotional decision-making 3) Provide smart, real-time feedback such as “You’re holding this trade longer than usual - consider reviewing your plan.” These features can be accomplished using techniques from Machine Learning such as Recurrent Neural Networks (RNN) as well as integrations of GenAI models for LLM interactions. This project blends machine learning, psychology, and finance, showing students how AI can be used to improve real human decision-making. Project Outcomes: - Design structured trade data models that capture trade features like Entry/exit, stop loss, strategy, confidence, and emotional state. - Build a remote data storage system (e,g, via PostgreSQL) - Compute key trading metrics like Profit/loss, holding time, reward-to-risk ratio, which would be used by the machine learning models to identify behavioral patterns, detect risky habits or inconsistencies. - Process user-written trade notes using NLP/LLM techniques. - Build an interactive journaling interface (Streamlit or Dash in Python) which convey Pre-trade guidance, In-trade alerts, and Post-trade feedback. - At the end of the program, the student will deliver a working web app prototype that stores and analyzes trades, learns trade performances/ patterns, and provides AI-generated feedback at different trade stages with documentation of their approaches, experiments, and findings.
AI Market Chart Pattern Recognition Engine App
Project Description: Teach AI to “see” and recognize patterns in real stock charts In this project, students will build a web-based AI platform that learns how to read stock market charts - just like a trader does. Users can choose a stock or crypto asset and a time scale (daily, weekly, or intraday), and the platform will generate interactive candlestick charts with volume and technical indicators layered on top. Students will then train AI models to recognize chart patterns and predict what might happen next. The AI models will attempt to forecast future price movements (for example: 1 hour ahead, 1 day ahead, or 1 week ahead). This project introduces students to: 1) Turning financial data into images AI can understand 2) Teaching computers to “see” patterns in charts using deep learning models like Convolutional Neural Networks (CNNs) and LSTMs, 3) Visualizing predictions in a clear, intuitive way, and 4) communicating the model findings using GenAI techniques (e.g., Visual Language Models). By the end, students will have built a real AI-powered pattern recognition engine and gained hands-on experience with deep learning in finance. Project Outcomes: - Build data pipelines for financial time series (OHLCV, indicators) - Transform market data into model-ready inputs (numerical + image-based) - Develop and test deep learning models (CNNs, LSTMs, or hybrid approaches) - Design prediction targets such as: Price direction (up/down/sideways), Strategy signals (trend-following, mean reversion), Visualizing predictions on interactive candlestick charts (Plotly), and integrating GenAI to generate human-readable feedback for predictions - At the end of the program, the student will deliver a working web app prototype that displays financial charts, generates model-based predictions, and provides clear, user-friendly prediction feedback with documentation of their approaches, experiments, and findings.