Justin Hatch

Designed an AI-driven decentralized hedge fund using Polymarket API, multi-agent LLM decision systems, and Solana-based staking. Built a six-agent Gemini debate system for market research, proposal evaluation, and autonomous trading.

Achieved 91% accuracy, enabling proactive churn detection. Performed data cleaning, feature engineering, EDA, and applied multiple ML models (Decision Trees, KNN, SVM, MLP). Improved performance with hyperparameter tuning and model validation. Authored a technical report covering methodology and performance analysis.

Implemented GBM–based Monte Carlo simulations to model stock price paths for prediction and scenario testing. Built a Streamlit dashboard for interactive experimentation, letting users choose tickers, set simulation parameters, and visualize expected prices, confidence intervals, and outcome distributions. Designed modular code for data ingestion, log-return computation, parameter estimation, and simulation to enable rapid iteration and future model extensions.

Analyzed Intel sustainability data to identify high-ROI improvements and presented actionable insights. Conducted a statistically rigorous A/B test to determine the best CTA button for increasing GRAMMY U membership. Delivered narrative-driven visualizations aligned with industry reporting standards. Completed as part of the Global Career Accelerator program.

Implemented linear regression using the Least Squares method. Read 100 data points from file, generated best fit line, visualized results with Matplotlib, and calculated prediction error.

Cleaned and prepared bank marketing campaign data for future analysis. Handled missing values, converted data types, reformatted dates, and structured output into relational CSV files for database import.

Produced high-quality screen-recorded demonstrations using OBS to train AI models on Microsoft Office workflows. Created detailed task walk-throughs with step-by-step reasoning and instructional clarity. Reviewed and annotated system-captured actions, providing structured “thought-process” explanations to improve LLM reasoning.