Advisor: Prof. Jorge Ortiz

Project: SmartBox - A Platform for In-Situ Sensing and Active Learning

JJ Slade Honors Thesis - August 2019-May 2020

• Created a full-scale platform to enable machine learning algorithm development and smart-building features such as occupancy counting and HVAC control in any building.

• Designed and implemented a scalable data collection platform with a front-end web application for data visualization and active learning, back-end web services for data processing, and large-scale data storage using TimescaleDB.

• Created a hardware prototype with 18 sensors for accurate data collection using Raspberry Pi and Arduino.

• Highest rated thesis project by a survey of faculty, selected to be 1 of 4 video presenters (attached below)

Project: AutoSeg: Changepoint Detection using Deep Autoencoders

October 2018 - June 2019

• Assisted development of autoencoder neural net-based algorithm for generalized segmentation of time series data.

• Designed, implemented and used a flexible testing framework in Python for assessing algorithmic efficiency and accuracy, allowing for multiple segmentation algorithms to be run across multiple datasets and assessed across custom metrics.

• Applied knowledge of fourier transforms to process and clean data and suggest improvements to segmentation algorithm.

• Created NLP-based fuzzy string matching tool to automate manual process of matching similar company names across spreadsheets in Python. Generalized tool for multiple applications across teams and wrote documentation.

• Automated stress testing process on team’s loan portfolio using Excel VBA. Eliminated manual copy-pasting across 4-step process through macros, manual reruns through iterating a scenario table. Generated formatted report as output.

• Created a systematic trade pitch based on NLP model, wrote daily world news briefs, and assisted live currency trading.