Elijah Qi

Hello!

I ♥ Coffee.

I'm currently pursuring Computer Science and Data Science double major at the University of Michigan.

I'm looking for an internship as a software engineer.

Dates	Work	Description
May 2022-Present	Research Assistant with Professor Kim at University of Michigan	Read all the json files including subfolders, parsed them according to the author’s name and the name of the published paper, assigned each article and author a different id, and merged them by comparing the similarity of their last name+first name’s first three letters+suffix strings to determine if authors need to merge their ids. Used dictionary in python to save authors’ name for every paper and used Union find to merge the authors whose name’s similarity(provided by difflib.SequenceMatcher()) is greater than 0.85. Improved efficiency of finding all the JSON files by finding one path of a JSON file and read one JSON file instead of saving all the paths and reading the JSON file from the beginning of the list, reducing the runtime from 43 minutes to about 2 minutes.
Jul.2021-Aug.2021	Industry Analyst Intern at The Pacific Securities	Retrieved the latest data on the number of stock prices and research reports in the automotive industry and saved them as JSON files by using Pandas and NumPy. Automated the data collection and collation including market value in circulation, latest trading day increase or decrease,reduced the data collecting time from 2 hours to 10 minutes.

Details
Developed an Instagram-like web application using React for front-end, REST APIs in Python using Flask as the back-end, and SQLite for database, supporting features include creating, updating, and deleting users, posts, comments, likes and infinite scrolling. Deployed this website using EC2 instance on Amazon Web Service. Encountered problem when handling number of likes and posting new comments. Solved by lifting up state to allow like-component and comments-component to share the data and prevent complex state management.

Details

Developed an Instagram-like web application using React for front-end, REST APIs in Python using Flask as the back-end, and SQLite for database, supporting features include creating, updating, and deleting users, posts, comments, likes and infinite scrolling.
Deployed this website using EC2 instance on Amazon Web Service.
Encountered problem when handling number of likes and posting new comments. Solved by lifting up state to allow like-component and comments-component to share the data and prevent complex state management.

Details
Designed and implemented a classifier to classify Piazza posts data according to the probability of the tag of the posts. Used the training data to figure out the percentage of posts of each type to the overall posts; figured out which type the post has the highest probability of belonging to based on the read-in data from the test set; assigned that type to the current post. Increased the efficiency of worst case from O(n^2) to O(nlog(n)) by implementing self-balancing Binary Search Tree to create a container called map to save the probability for each tag and word.

Details

Designed and implemented a classifier to classify Piazza posts data according to the probability of the tag of the posts.
Used the training data to figure out the percentage of posts of each type to the overall posts; figured out which type the post has the highest probability of belonging to based on the read-in data from the test set; assigned that type to the current post.
Increased the efficiency of worst case from O(n^2) to O(nlog(n)) by implementing self-balancing Binary Search Tree to create a container called map to save the probability for each tag and word.

Details
Designed and implemented an automated web screenshot software that scrolls through the first two pages of a search engine and generates screenshots based on a given company name and corporate name. Opened the search engine type in the search content using Selenium; found whether the existence of the next page in the CSS selector; saved at most two pages of results according to the CSS selector result. Used JavaScript to get the length and width of the page and then take a screenshot, loop every 1080PX to cut a picture, and then named it according to the order of precedence. Reduced the preparation work for an enterprise risk assessment to 2 hours on average.

Details

Designed and implemented an automated web screenshot software that scrolls through the first two pages of a search engine and generates screenshots based on a given company name and corporate name.
Opened the search engine type in the search content using Selenium; found whether the existence of the next page in the CSS selector; saved at most two pages of results according to the CSS selector result. Used JavaScript to get the length and width of the page and then take a screenshot, loop every 1080PX to cut a picture, and then named it according to the order of precedence.
Reduced the preparation work for an enterprise risk assessment to 2 hours on average.