Portfolio
Automated Website Building Assistant
The Client
A web development agency that was creating a web development plugin for WordPress
The Solution
A GPT-powered AI assistant. The assistant is capable of querying the tool’s documentation and answering any relevant questions. It also has additional tailored features for helping new users become comfortable with the plugin. More details will be added once the tool has been released.
The Tools
OpenAI’s API, Google Cloud Platform, ChromaDB Vector Database, FastAPI, Docker

AI Powered Insights for Esports
The Client
A startup that was developing an esports coaching platform
The Solution
A computer vision powered tool that extracted information and statistics from raw gameplay footage. The solution included model training from scratch, containerization, and deployment. The tool was trained for two games (Super Smash Brothers Ultimate and Overwatch) which do not have open APIs, so all the statistics needed to be extracted from gameplay footage alone. The tool was deployed on Google Cloud Platform and capable of supporting a nearly unlimited number of concurrent users.
The Tools
Tensorflow, Google Cloud Platform, MongoDB, Flask, Docker

Feature Engineering and Model Training for Unique Data
The Client
A security company that was creating an entirely new “smart metal detector”
The Solution
Trained and deployed custom-built models for an entirely unique set of data that consisted of raw electro-magnetic sensor data. The solution included consulting with physicists to determine the best features for the unique data and developing the entire ML-Pipeline.
The Tools
Pytorch, MLFlow, Triton Inference Server, Nvidia Jetson, Docker

Facial Recognition to Remind Users During the Pandemic
The Client
A tech company that required an in-person presence during the pandemic
The Solution
A tool that connected to the security camera’s API and to Slack to remind users to log their symptoms upon arrival at the office. The tool required users to upload a picture of their face so that it could calculate the embeddings and recognize the user when they arrive in the office. The tool was opt-in only and was developed due to many employees forgetting to log their symptoms regularly.
The Tools
Pytorch, FastAPI, Dash, Slack Bot API
