HI, I'M JOEL

ML/CV Engineer with end-to-end experience

Deep learning + classical CV, strong debugging instincts, and a systems background that helps ideas come to life — from data and training to GPU/edge deployment.

A portrait of Joel Wigton, a machine learning engineer

ML/CV & Systems Skills

Tools I use to build, train, and deploy vision models.

I spent 15 years building systems across hardware, firmware, and software — and now apply that same engineering rigor to machine learning and computer vision applications.

My superpower is problem solving: learning fast, researching deeply, and getting to the truth of the problem.

AI and ML are superpowers for problem solving as well — and combined with my strengths, they become a powerful force multiplier.

ML/CV Projects

NLP Projects

Generating Amazon Book Reviews with Transformers

Problem: Generate realistic product-review text from a small dataset.

Approach: Fine-tuned a GPT-style causal language model (DistilGPT2) with Hugging Face tooling.

Result: Produced coherent, style-consistent synthetic reviews suitable for demos and augmentation experiments.

Make Your Own CliffNotes (Summarizing Text)

Problem: Summarize long text into useful short notes.

Approach: Fine-tuned a T5 summarizer and evaluated with ROUGE plus qualitative checks.

Result: Generated readable summaries with measurable improvements over baseline settings.

Answering Questions from Product Reviews

Problem: Extract direct answers from noisy customer review text.

Approach: Built a transformer-based QA pipeline with preprocessing and evaluation.

Result: Delivered a working QA system that returns grounded answers from review context.

Ingredient Standardization via Machine Translation

Problem: Normalize messy ingredient text into a consistent schema.

Approach: Trained a seq2seq translation-style model using synthetic paired data.

Result: Improved consistency of ingredient outputs and reduced manual cleanup effort.

What Kaggle WON'T Tell You About Your Notebooks (Clickbait Detector)

Problem: Detect misleading or clickbait-style titles.

Approach: Compared models (RNNs/transformers) and tracked precision/recall/F1 with threshold tuning.

Result: Built an end-to-end text classification workflow with interpretable tradeoffs.

Detecting Android Malware from App Permissions

Problem: Malware detection is skewed/imbalanced and accuracy can lie.

Approach: Trained tree-based models and optimized using precision–recall curves instead of accuracy.

Result: Improved detection quality under class imbalance and built a deployable classifier baseline.

Featured Web App

A web product I built, launched, and grew to 45K+ weekly active users, including hundreds of paid subscribers.

Shipped & Adopted

Built, launched, and grew a real product used by tens of thousands weekly.

Operated in Production

Kept it reliable at scale: monitoring, iteration, and continuous improvements based on real user behavior.

Product + Business Execution

Took it from idea → distribution → monetization, including a freemium subscription model with paying customers.

Hardware, my first love

An electrical engineer by schooling, I will always have a special place in my heart for hardware. These projects are a snapshot of the “build it for real” side of me.

C-Rex

Converted a Honda CRX into an EV.

The open hood of a homemade Honda CRX electric car

BMMPR

Built an internet-connected car alarm.

A screenshot from sitesunseen.co, a landing page design by Joel Wigton

Stunly

Prototyped a wrist-worn, motion-activated self-defense device.

A screenshot from plaidish.com, a landing page design by Joel Wigton

Curb

Built a fridge device to curb snacking habits.

College

University of Michigan, Ann Arbor

Masters, Electrical Engineering.
Summa Cum Laude

Bachelors, Electrical Engineering.
Magna Cum Laude

Graduated with Honors
Dean's List
Two senior class "Best Project" Awards

Certifications & Coursework

Hands-on training in deep learning, computer vision, NLP, and production-grade software fundamentals.

TensorFlow Developer Certificate

Practical TensorFlow skills: building, training, and evaluating real models.

TensorFlow Keras

Udacity's Computer Vision Nanodegree (PyTorch)

Core CV workflows: feature extraction, deep models, and applied perception projects.

PyTorch Facial Keypoint Detection Image Captioning SLAM

DeepLearning.AI's Natural Language Processing Specialization

From classical NLP to transformers: embeddings, seq2seq, attention, and evaluation.

Transformers Seq2Seq Machine Translation

DeepLearning.AI's Deep Learning Specialization

Neural nets, CNNs, sequence models, and training best practices.

Face Recognition YOLO Image Segmentation Sentiment Analysis Neural Machine Translation

Stanford's Machine Learning Specialization

Strong foundation in supervised learning, model selection, and practical ML.

Anomaly detection Movie recommendation engine RL Lunar Lander

They said it, not me..

“It is without hesitation that I give Joel my highest praise.”

“Joel performed well beyond my high expectations. He was self-motivated, and was quickly able to set his own goals and timelines. He incorporated innovations to meet demanding project requirements and delivered across productivity, initiative, responsiveness, and professionalism.”

Thomas Grutkowski

Intel Principal Engineer

“We still talk about some of those early days where you really saved our bacon”

Thomas Reilly

CEO at AST

“Managers and technical leads knew they could trust him”

“Throughout his time on my team, Joel showed a knack for learning new subjects quickly, he showed excellent communication/teamwork, and he reliably met his schedules. I’d be happy to have Joel return to my team!”

Paul Thayer

Intel Director

“Passionate, precise and creative”

“Joel understands complex matters even when outside of his direct area of expertise. He is quickly able to learn new software tools. Joel’s work ethic and attitude are an asset to any team.”

Mike Barber

Principal Engineer at Edge Scientific

[email protected]

GitHub

Kaggle