Jixuan Wang

Fullstack on model pipeline 1) Developed Feature Store to calculate and store real-time and offline features for prediction using data pipelines (Kafka, MQ) and databases (Redis, Hive). 2) Trained a series of fraud detection models in FinTech. Slashed the fraud loss significantly. 3) Developed model deployment framework in clusters, support different types of model (Spark, Sklearn, PMML, pb, checkpoint). 4) Developed automatic model performance assessment tools, including AB testing. Data Collection 1) Developed data crawling tools to collect public opinion. Analyzed data to detect potential malicious users and attacks. 2) Crawled data from multiple resource. Focused algorithms 1) Developed address similarity detection algorithms. Use NLP algorithms to analyze non-formatted addresses, compare geographical similarity of these addresses. 2) Developed model assessment algorithms.

Small business loan assessment.