My Portfolio

About Me

I am a graduate student studying Data Science at the University of Michigan School of Information. I am interested in data mining, machine learning, deep Learning, and NLP. Last summer, I was a data science intern at UHG. I worked on uplift modeling to predict the effect of marketing campaigns on each individual customer and provided actionable insights to fuel growth for the comapny. Currently, I am looking for full-time data science positions starting Jan 2022.

Experience

UnitedHealth Group

Data Science Intern | San Francisco

Conducted an end-to-end machine learning and causal inference model on observational data to predict the effectiveness of the marketing campaigns on each individual customer (A/B Test)

performed feature selection on over 1000 sparse features using f-statistics; compared two model approach, class transformation, solo models and selected the best performing model with AUC 0.78

Improved the campaign effectiveness by targeting only the top 20% of users

proposed the end-to-end productization plan based on MLflow to marketing stakeholders; publishing the model API to the downstream consumers

Education

University of Michinga

Jan 2020 - Dec 2021

Master of Science in Information | Data Scince Track

Coursework: Computational DS & ML, Machine Learning, Natural Language Processing, Deep Learning, Big Data Analytics, Data Mining, SQL & Databases, Applied Machine Learning, Data Manipulation and Analysis, Information Visualization, Databases and Application Design

Alzahra University

Jan 2016 - Dec 2018

Master of Business Administration (MBA)

Coursework: Statistical Analysis Using Python, Management Information Systems, Data Mining, System Dynamics

Iran University of Science and Technology (IUST)

Sep 2008 - May 2012

Mechanical Engineering

Projects

A Vertical Search Engine for Airplane Crashes

In this project, I designed and built a vertical search engine using Python. I used Beautiful Soup library to scrape the data and designed SQL databases to store them. The User can select a year, and decide the sorting method to see various aggregations and charts about the happened crashes in that year. Moreover, I Used TF-IDF metric and bm25 function to retrieve and rank the most relevant documents for each query terms. Finally, I implemented the graphical user interface using Python and Flask.

Project Repository | Article | Video

Predicting Individuals’ Willingness to Pool in a Ride-Hailing Trip

In this project, I developed, trained, and tuned multiple machine learning models (LR, GB, RF, ADA) to predict the willingness of individuals to share a trip using a novel dataset of 18.3 M records from ridesourcing trips achieving f-1 score of 80%

EDA_Notebook | ML_Notebook | Report

Song Lyrics Generation with Artificial Intelligence

This project is part of Natural Language Processing Class. I fine-tuned a transformer language model (GPT-2) to generate lyrics. In the next step I built a classification model using BERT to detect whether a song is machine-generated or not with 85% accuracy.

3D Object Detection for Autonomous Vehicles

This is the final project for the Big Data Analytics class. I was the team leader for this project. We used Lyft Perception Datesent Kit to build models to percept 2d and 3d objects around an AV. For the 2D part, We used YOLO algorithm on 100 GB 2D images. For the 3D part, we trained U-Net algorithm on 100 GB 3D-point-clouds to percept 3D objects with MAP of 0.51.

View Project

Temperature Predicton for Trout Lake

This project is part of Applied Machine Learning class. In this kaggle competition, I led a team to train and tune mulitple ML models to predict the temperature of Trout lake for one year. We trained a linear regression algorithm with feature expansion, which reached an MSE of 0.022. My team ranked third in this kaggle competition.

View Leaderboard

Elham Amini

Data Scientist