Kate Lassiter's repositories
Cornell-Machine-Learning
Practice from My Machine Learning Certificate from Cornell
Machine-Learning
Machine Learning in Scikit-Learn and TensorFlow
Natural-Language-Processing
CNN, LSTM, GPT and other NLP models with PyTorch and Transformers
Repeat-Sales-Model-using-Distributed-Computing
Forecasting corporate bond returns with a repeat sales model on 72 distinct billion item matrices ~ 1TB.
Web-Scraping-SEC-Data
Web scraping commercial paper and negotiable certificates of deposit data from the SEC EDGAR public website https://www.sec.gov/edgar/searchedgar/legacy/companysearch.html
Bayesian-Machine-Learning
Deriving expectation maximization and variational inference algorithms from scratch.
cc22tt
Source files for community contribution repo for EDAV Fall 2022 Tues/Thurs
media_disinformation
Data visualizations exploring media disinformation, including interactive D3 javascript
Misc-Columbia-Projects
Some projects from my Columbia University data science engineering program
PolicyPlotter-User-Interface-for-Plotting-Board-Charts
Policy Plotter – R Shiny Application: Solved a core problem facing the Federal Reserve Board by independently developing a plotting application to vastly increase the efficiency of creating charts/exhibits for The Federal Open Market Committee (FOMC). This committee determines monetary policy for the entire United States. I automated the process of chart creation which draws focus away from data and code integrity.
Space-Optimized-Computer-Vision
A space optimized CNN developed through synchronous distributed training, weight pruning, and quantization in Vertex AI on GCP
String-Similarity-Custom-Functions
Text Classification: Fast, custom string similarity functions in Python mapping lambda functions to NumPy arrays, assigning issuers to one of 10,000 classes.
Umass-Senior-Project-2019
The purpose of my application was to solve a problem many businesses (small businesses in particular) face. They do not know how much to produce, where to price, how much to spend on advertising and many other questions. Eden’s purpose was to answer these questions for them easily and with no technical acumen required by the user. Eden would model supply and demand equations using ordinary least squares (OLS) regression on the user’s data to form the best fitting supply and demand equations possible. The best fit was to be ensured by regressing each variable against demand or supply, determine the best shape via the highest adjusted R2, and then do an OLS regression and simplistically tell the user what the results mean. Eden would attempt multiple shapes like linear, logarithmic, cubic, quadratic, and inverse. The user interface would be easy to navigate and user-friendly.