Adam Liu

I participated in developing this app during my internship at Ailink Technology Cooperation.As my first ever internship, I learned to apply theoretical knowledge like data structures,networking, and algorithms on a real project.I started with small features, like checking if a password is valid on the signup page, or if an email is already registered. After getting more comfortable with the technology stack, I moved on to more complex ones like the pagination feature in the homepage. In my opinion, one of the most important takeaways I got from developing this project is to never be afraid of not knowing new technology, be it new programming languages or new frameworks, because software engineering is an ever-evolving discipline. As long as you are always eager to learn, don't be scared by the technology you don't understand (yet!).This app is built using React Native/Redux and AWS.

This is a machine learning project done as an assignment for CSE158,Recommender system and web mining at UCSD. I implemented a similarity-based/Bayesian Personal Ranking recommendationsystem for Steam games in Python. The dataset is directly crawled from Steam, containing anonymized informationon over 10,000 users and their ratings (like/dislike) on games. Using only Jacobian/cosinesimilarity with some hyperparameter tuning, I was able to achieve over 60% accuracy in predicting whether people will like a specific game. Adding in Bayesian Personal Ranking, the accuracy approached 80%, which is at the top 5% in the class. From developing this project, I learned about the complete machine learning pipeline, from data collection to prediction. This is when I also realized that machine learning and AI are not just about programming. To make sense of a large dataset, I also needed knowledge in statistics, calculus, and other domain specific fields.

This is a text mining group project that I am a part of. As the name suggests, we looked at over 40,000 of Trump'stweet and analyzed its sentiment using methods like LSTM, Tf-IDF, and other classic supervised machine learningalgorithms. For each tweet, we calculate a sentiment score based on word tokens. Then, using the score astraining data, we ran various algorithms on the dataset to test accuracy. The major difference between thisproject and my previous ones is that we were also asked to write an academic paper describing our methods,results, and conclusions. I was tasked with running and comparing the performance of different supervised algorithms on our dataset. Since other members of our group were all experienced academic writers, this project introduced me to the methods of academic writing.