Dog breed identification
Used transfer learning to identify the breed of a dog, or find the dog that looks like you
- Tools used: Python, opencv, AWS sagemaker
- Category: Clasification, Deep learning
- Year: Apr 2020
Affinity propagation clustering package
Built a Python package that implements affinity propagation from scatch - A technique based on message exchange between data points and do not require pre-specifying the number of clusters.
- Tools used: Python, TestPyPi
- Category: Clustering, Teamwork, Software engineering
- Year: Apr 2020
Plagiarism detector
Using NLP attributes such as containments for feature engineering and Naive Bayes classification for detecting plagiarism.
- Tools used: Python, AWS sagemaker, AWS Lambda and Amazon API Gateway
- Category: Classification, Machine learning deployment
- Year: Apr 2020
Machine learning interpretation app
An app to automatically interpret blackbox high-performance machine learning models
- Tools used: Python (tree ensembles, Streamlit, SHAP, ELI5, Altair), Heroku, Docker
- Category: ML interpretability
- Year: Mar 2020
Predict patient survival 24hrs after admitting to ICU
Evaluated different boosting algorithms on a dataset including features from various vitals and lab test results. Achieved 0.905 AUC
- Tools used: Python (catBoost, XGBoost, lightGBM, SHAP)
- Category: Supervised Learning
- Year: Feb 2020
Flask web app for text summarization
Deployed a flask app with continuous delivery that performs extractive summarization on long text or lists of web articles from user input
- Tools used: Python (Flask, Gensim, Transformers), GCP (App Engine, CloudBuild)
- Category: NLP, App development, CI/CD
- Year: Jan 2020
Epidemic meets social media
Identify the different concerns among pro- and anti- vaccine group surrounding #flu and #flushot from Twitter, visualized the interactions between the groups and the influencers within them
- Tools used: R (rtweet, tidyverse, ggraph)
- Category: Exploratory Analysis, NLP, Network Analysis
- Year: 2018
Customer segmentation
Modeled user the segments of a credit company through K-means clustering and t-SNE. Made recommendation on different targeting strategies
- Tools used: R (K-means, t-SNE), Tableau
- Category: Supervised learning
- Year: 2018
