Liu Liu

Implemented a Heart Drug Target Detector to identify drug targets from 1.4 billion candidates, achieving a precision score of 0.91.

Led a team of 3 data scientists to extract data from 8 million articles, using NLP and LLM techniques for feature generation. Applied feature selection with Boruta and mRMR, trained models (Logistic Regression, Random Forest, XGBoost, SVM, CNN), and deployed the model as a Flask app on AWS.

Developed an AI-powered end-to-end MVP that automates paper selection, content extraction, and generation, reducing research time by 100 hours per project.

Collaborated with a cross-functional team to implement RAG with Chroma vector databases and GPT-4. Developed a Flask-based backend and API, coordinated with front-end and product manager teams, and deployed the application on Azure.

Developed a Heart Attack Detector using health data, achieving a recall score of 0.81.

To identify high-risk individuals for heart attacks, I integrated data from a local hospital database and web scraping. Derived features using historical survival analysis models and trained multiple classification models, including KNN, XGBoost, and Artificial Neural Networks.

To identify gene targets promoting heart regeneration, I developed a bioinformatics data analysis pipeline. Collected and processed RNA-seq data, performed Quality Control, Read Mapping, and Quantification. Applied bioinformatics analysis to extract insights, leading to the discovery of 47 heart repair genes.