• Developed a search engine that provides related support articles for users' questions.•Maintained an NLP pipeline that periodically tags new articles and updates the index storage system on Redis.• Designed a query processing system that consists of a MapReduce-based retrieval module that collects relevant articles and a tf-idf ranking module; This system was back-tested on historical user logs using MRR, MAP, and NDCG.

• Developed an anomaly detector and a natural language processing pipeline for a financial software company that provides accounting and tax preparation software for small businesses, accountants, and • Developed an SVM-based anomaly detector using PyOD API that filters incoming HTTP requests from client sites which protects the server from malignant requests such as DoS attacks; the proposed anomaly detector achieved 91.2% accuracy in offline testing using service logs on a 90-day window.• Designed and implemented an NLP model using spaCy and scikit-learn API that automatically classifies the expense category for receipts; the proposed model utilizes k-means algorithm to cluster expense description based on semantic similarity calculated from embedding vectors; evaluated on a manually labeled test set, this model achieved 80.2% top-1 accuracy and 91.3% top-3 accuracy on 3,562 expense categories.