Pooker: A Not So Shitty Poker AI

Highly modular C++ project built culmulatively on git with over 100 commits over 2 months.

Uses Monte Carlo Tree Search to determine most profitable action, simulating over 7 million outcomes each decision

Properly allocated dynamic memory over 500,000 allocs without any memory leaks

Live Free

PennApps XIII Winner! Predicts diabetic patients future blood sugar levels using artificial Neural Networks. Implemented stochastic, random-batch gradient descent for 91% predictive accuracy!

Robot Racing Traffic Detection

Using MATLAB to prototype traffic detection algorithmns, using hsv and rgb filters to detect green go light. Porting code to work in C++ using OpenCV.

Step Up

An Android pedometer that uses accelerometer readings to count steps. Invovles use of low-pass filter to reduce noise and finite state machines to discriminate between steps and non-steps.

I'm planning to train a neural network to handle the step detection instead of the finite state machine for higher accuracy.