William Yang

Software Engineer Intern

• Worked on developer productivity team to speed up user requests by 3 times for data engineering AWS machines for 700 engineers
• Wrote cron jobs in Puppet/Ruby that maintained standby pools of 300 remote hosts for GPU and high compute data machines
• Enabled instance storage caching on data engineering remote machines, speeding up git operations and build times by 5 times
• Created update workflows to maintain and update packages/repos for a pool of 250 EBS and attach them to remote hosts on user request
• Configured Grafana dashboards and 10 alarms to monitor workflow, cron jobs, and remote machine health data, ensuring reliability

Software Development Engineer Intern

• Collaborated with multiple Amazon Stores teams to research and build a SQL database in AWS Open Search for 90 thousand fashionproduct items. I also implemented a 10ms latency data stream pipeline in AWS Lambda to update the database with new products items.
• Conducted 5 technical demos to improve database features and presented to Amazon Fashion Tech directors.
• Reduced search latency times in the fashion products internal database by ten-fold through optimizing AWS tooling.
• Fully integrated a customer feedback and review form for fashion products on the Amazon Stores site with RestAPI backend.

Software Developer

• Built data ingestion pipeline that ran in Docker and stored 10TB worth of Kafka message queue alerts to a PostgresSQL database.
• Setup a live integration testing framework in GitHub CLI using 5 docker containers to simulate critical system components.
• Built a FlaskApp website with ReactJS displaying received records and email alert 3 thousand users when neutrino events are detected.

Software Engineer Intern

• Created a Big Query database pipeline with SQL to analyze data over 1 million customer cell line quality or drop rates.
• Implemented an ensemble of machine learning models in PyTorch such as XGBoost, Random Forest Classifiers, and Logistic Regression to identify over ten thousand Verizon customers at risk of switching to competitors.
• Designed algorithms for fine-tuning models to customer data from 20 different US regions and demographics.
• Continuously worked in code review, load/stress testing, integration/unit/Blackbox testing.

Machine Learning Engineer Intern

• Led the intern team in writing a paper summarizing our model's results. Developed an ensemble model version of BERT on AWS to perform question answer tasks on various documents from legal, medical, and academic settings, beating baseline models by 10 percent in accuracy.