Integrated ALRITE (Android app for efficient diagnosis of pneumonia) with OpenMRS(medical record platform).item Designed and implemented a new front-end of Alrite for better UI and compatibility with OpenMRS system. Utilized OpenMRS as patients' health condition storage and tracker, support for history check and wider analysis. item Our work was widely deployed and used in 170+ clinics in Uganda, and tracing health condition for 2k+ patients.

Developed flexible cloud based web-service, integrated \& visualized internal records for company analyze.Designed efficient project with overall Architecture and Implementation Strategy, focusing on User experience.Full Stack development, familiarized and wisely applied multiple Internal Services with best code quality.With my work, customers (150+) could reduce 5-8 hours per week for data integration and analysis.

Made use of TypeScript and React environment to build navigation tool for UW campus.Applied Dijkstra Algorithm with Json dataset to draw the shortest path from two positions.Output the distance and directions to guide the customers.

• Applied Bayes' Theorem in NLP Natural Proof generation to support higher generation accuracy.• Dived into Seq2seq models such as Bart and GPT2, developed Prior, Channel Model, and Testing strategy.• Made the utility of the prior model and achieved correct proof pattern generation with mathematical step error.• Current moved to Proof Judgement, training model to identify error categories of Natural proofs.

• Quick Learned knowledge about NLP learning strategies and NLP models.• Filtered data and using data set to netune Bert, Bart, Roberta, T5 generate prediction for Mask LanguageModeling with numbers.• Designed self-generation of informative sentences, enhancing the prediction accuracy by around 30%.

• Assisted in developing private algorithms and training models for optimizing and prediction.• Designed algorithm for railway repairing workow, use linear programming to nd least occupation.• Collected custom chemical inputs, use Historical Optimization and reduce power plants pollution by 21%.

Developed a customer post-board, including account creation, login in, and like systemEnabled yipper search with user-name based on SQL queries, displayed results based on like numbers.Set up the API with node.js, enabled remote connection, and supplied yipper data.