I'm Joel Kim.

I'm a junior at Northwestern University studying Computer Science and Data Science, and I'm passionate about machine learning, software engineering, and all things data.

This website is a little peek into my work life. Enjoy!

My Projects

Spotify Playlist Quiz

A full-stack web app for a ”Guess the Song” quiz dynamically generated based on the user’s account.

REST APIs Flask MongoDB

GitHub

Airbnb Price Predictor

A machine learning model to predict Airbnb prices using pandas and scikit-learn to reduce RMSE by over 60%. The final model employed stacking ensembling with gradient boosted models and a RidgeCV meta-model to optimize predictive performance.

Pandas Machine Learning Scikit-learn

GitHub

Digit Recognizer

Three neural networks using NumPy, PyTorch, then ResNet-18, achieving 94%, 97%, and 99% accuracies.

PyTorch Neural Network

GitHub

Sports Betting

Sports betting, or sports trading, is a fascinating and complex industry. On one hand, it’s a predatory market where sportsbooks profit off the losses of the average bettor. On the other, it's a market where those who approach it analytically can find inefficiencies and edges.

Over the past three years, I’ve gone on a long journey with sports betting. I’ve tested countless strategies, conducted extensive research, and engaged in discussions with peers, all while wagering tens of thousands of dollars.

I could talk about it for hours, but below is a view into my profit/loss over time, and some of the strategies I've used over the years.

A snapshot of my profit/loss trends over time.

Strategies I've employed:

Tailing Cappers: The starting point for many bettors copying bets from others, extremely dependant on who the bets come from.
News Updates: Placing bets based on anticipated news events or reacting quickly to breaking news to capitalize on market shifts, moderate variance.
Arbitrage: Exploiting pricing discrepancies between sportsbooks to generate risk-free profits, no real variance.
Positive Expected Value (+EV): Placing bets where statistical probabilities favor long-term profitability, moderate variance.
Correlated Parlays: Constructing multi-leg bets with interrelated events to maximize hidden value, higher variance.