I have completed several personal software engineering projects. My goal was to solve everyday problems with data science techniques. I built the algorithms, architected the databases, designed the user interfaces, and built tests 100% independently, full stack.
I created an iOS app with Swift that generates unique ordered playlists from one input song.
I developed a machine learning model to dynamically generate optimally ordered playlists in Θ(log(n)) runtime from an input song. I Authored 100% of code from scratch, including APIs, libraries, and the machine learning model. I architected a KD-Tree variant that optimized and maximized efficiency. In my variant of the KD-Tree, I overcame the KD-Tree curse of dimensionality and improved runtime from Θ(n) to Θ(log(n)).
I structured & optimized an original database of over 800,000 songs with 14 predictive song features and attributes, and engineered a Google Cloud database allowing me to deploy a new model remotely without the need for an App Store update. I also designed and implemented a referral-tracking system and complex app analytics to improve conversion and increase the number of users. I also fully integrated my app with the Spotify API for a seamless user experience.
I created an iOS app with Swift that uses machine learning to detect counterfeit sneakers. It is currently on the iOS App Store.
I independently created a training dataset of over 10,000 images, each being labeled as either counterfeit or authentic. I then built a custom Tensorflow machine learning model and trained it on my input data. I was able to attain 85% accuracy in identifying counterfeit sneakers with the model--higher than most humans are capable of.
I launched Authentic8r in late 2018, and since launching, I have attained 1,300 total users. I have collected over 20,000 images from users, enabling me to improve the accuracy of the model further. There are currently 100 weekly active users of Authentic8r.