Skip to content
View JetHayes's full-sized avatar

Block or report JetHayes

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
JetHayes/README.md

Hello World, I'm John Cavanaugh

I'm a PhD candidate in Aerospace Engineering at the University of Cincinnat at the AI BIO Lab focused on fuzzy logic, evolutionary algorithms, AI architecture, and explainable AI. I also run the Plunk Foundation, building privacy-preserving coordination systems for organizations serving vulnerable families globally.

I build AI systems with transparency, explainability, and privacy by design


What I Work on

  • Fuzzy inference systems — Mamdani, TSK, and GA-optimized FIS built from scratch
  • Evolutionary algorithms — genetic algorithms for automated system tuning and combinatorial optimization
  • Explainable AI — interpretable models for high-stakes domains
  • Privacy-preserving ML — fuzzy feature augmentation as a privacy layer under differential privacy

Featured Projects

A fuzzy inference system where membership function parameters and rule outputs are evolved by a custom genetic algorithm. Built from scratch in Python. Best training RMSE: 0.0373 after 13 trials of systematic hyperparameter exploration. Python fuzzy logic genetic algorithms from scratch

Finds the minimum spanning tree for a 9-node offshore pipeline network using a genetic algorithm with Prüfer sequence encoding. Achieves optimal total pipeline length of 41. No graph optimization libraries. Python graph theory combinatorial optimization Prüfer sequences

First-order TSK fuzzy inference system trained via gradient descent on hydroelectric power plant data. 25 learned rules, 75 parameters, built entirely from scratch. Final RMSE: 17.46. Python fuzzy logic gradient descent from scratch

Full ML pipeline on IRS nonprofit data — regression, classification, and a KMeans hybrid clustering approach. MLP achieves R²=0.82. Built with scikit-learn on a real-world messy dataset. Python scikit-learn regression classification clustering

Fuzzy c-means membership values as noise-resilient features under differential privacy. MLP improves +4.74% RMSE. Presented at NAFIPS 2026. Co-authored with Tri Nguyen and Dr. Kelly Cohen. Python privacy fuzzy logic NAFIPS 2026 published

Mamdani FIS for intelligent prioritization of student peer review feedback across frequency, sentiment, and detail. Reduces HIGH priority inflation by 17%. Presented at NAFIPS 2026. Python fuzzy logic NLP education NAFIPS 2026 published


Background

Before the PhD I spent 13+ years in industry and founded a student platform company. I turned down a significant acquisition offer rather than build surveillance technology. That decision still anchors how I think about what I build and why.


Stack

Python PyTorch MATLAB NumPy scikit-learn


Find Me

Popular repositories Loading

  1. JetHayes JetHayes Public

    Hello world, this is my profile

  2. ga-pipeline-network-optimization ga-pipeline-network-optimization Public

    Minimum spanning tree for an offshore pipeline network using a Genetic Algorithm with Prüfer sequence encoding. Built from scratch in Python.

    Python

  3. nonprofit-revenue-prediction nonprofit-revenue-prediction Public

    ML pipeline predicting US nonprofit revenue from IRS data. Regression, classification, and KMeans hybrid clustering using scikit-learn. Built from scratch on real-world messy data.

    Python

  4. ga-fuzzy-inference-system ga-fuzzy-inference-system Public

    GA-optimized Fuzzy Inference System built from scratch in Python. Evolves membership function parameters and rule outputs to approximate sin(x)·cos(y). Best RMSE 0.0373. No fuzzy libraries.

    Python

  5. tsk-fuzzy-system-gradient-descent tsk-fuzzy-system-gradient-descent Public

    First-order TSK Fuzzy Inference System trained via gradient descent on hydroelectric power plant data. 25 learned rules. Built from scratch in Python. No fuzzy libraries.

    Python

  6. fuzzy-peer-feedback-prioritization fuzzy-peer-feedback-prioritization Public

    Mamdani fuzzy inference system for intelligent prioritization of student peer review feedback across frequency, sentiment, and detail. Reduces HIGH priority inflation by 17%. NAFIPS 2026.