Project: AI-Assisted Application Refactoring for App Modernization (on-going)- Implemented a novel hierarchical clustering based algorithm to decompose monolithic applications into microservices based on run-time traces and static code analysis (product Beta release, 1 paper, 2 patents).- Built a suite of evaluation metrics to assess the quality of recommended micro-service candidates.Project: AI Powered Cloud Migration- Developed an Active Learning microservice which… Show more Project: AI-Assisted Application Refactoring for App Modernization (on-going)- Implemented a novel hierarchical clustering based algorithm to decompose monolithic applications into microservices based on run-time traces and static code analysis (product Beta release, 1 paper, 2 patents).- Built a suite of evaluation metrics to assess the quality of recommended micro-service candidates.Project: AI Powered Cloud Migration- Developed an Active Learning microservice which simultaneously optimized data labeling and model performance by identifying and submitting high impact unlabeled data points for manual labeling. - Proposed a novel query strategy to deal with concept drift which improved accuracy by 4% (1 paper, 1 patent).- Built machine learning models to classify migration approaches for workloads being moved to the Cloud. - Decreased the time necessary to migrate workloads to the cloud by up to 17% by creating a STRIPS based planning engine that automatically generated customized steps suitable to each workload's characteristics (1 paper, 1 patent). Show less

Project: Sepsis Early Prediction Support Implementation System- Proposed a novel deep learning framework based on CNN and LSTM to perform early prediction of septic shock (3 papers).- Improved the prediction result by 7% (F1 score) compared to the state-of-the-art LSTM model.- Applied deconvolutional neural networks to visualize the different stages of sepsis by rendering the severity of each stage using realistic human faces.Project: Educational Data Mining for… Show more Project: Sepsis Early Prediction Support Implementation System- Proposed a novel deep learning framework based on CNN and LSTM to perform early prediction of septic shock (3 papers).- Improved the prediction result by 7% (F1 score) compared to the state-of-the-art LSTM model.- Applied deconvolutional neural networks to visualize the different stages of sepsis by rendering the severity of each stage using realistic human faces.Project: Educational Data Mining for Individualized Instruction in STEM Learning Environments- Applied deep learning/probabilistic graphical models to track students' knowledge using logs from tutoring systems. - Improved the accuracy of student’s knowledge level estimation by 10% compared to state-of-the-art approaches (3 paper). Show less

Project: API Discovery using Unsupervised Approach- Proposed a general framework consisting of techniques such as topic modeling, deep learning, and community detection to recommend API's by matching the intents of users Natural Language queries to API descriptions (1 paper, 1 patent).- Developed a web interface using Flask which allowed users to search for API's and provide feedback rating the relevance of suggested API's.

Project: Feature Engineering for Sepsis Early Prediction- Performed data standardization for structured/unstructured data and extracted features for downstream modeling tasks. - Applied an ensemble machine learning approach to both predict sepsis patient outcome using electronic health records and determine which comorbidities and pre-existing conditions contributed most to disease progression.

- Applied machine learning models to predict the likelihood of users terminating their licenses to boost customer retention.- Performed complex queries in Postgres SQL and Amazon Redshift for feature engineering. - Worked in an AGILE software team adopting scrum development methodology, utilizing JIRA to plan sprints, distribute tasks and monitor progress, and using Bitbucket for version control.

- Conducted interdisciplinary research with a team of 10 members from diverse engineering backgrounds.- Conducted genome analysis on E.Coli with more than 10 million bases.- Utilized Python packages "Matplotlib" to visualize ribosomal footprint during protein translation.- Proposed a new index to predict protein yield with significant higher accuracy (nearly 25% improvement in R^2).- Implemented dynamic programming algorithms in Python to optimize gene sequences for higher… Show more - Conducted interdisciplinary research with a team of 10 members from diverse engineering backgrounds.- Conducted genome analysis on E.Coli with more than 10 million bases.- Utilized Python packages "Matplotlib" to visualize ribosomal footprint during protein translation.- Proposed a new index to predict protein yield with significant higher accuracy (nearly 25% improvement in R^2).- Implemented dynamic programming algorithms in Python to optimize gene sequences for higher protein yield. Deployed the application to the RiboLab official website with thousands of potential users.- Worked through Unix command line programming via SSH on linux servers running Ubuntu. Used Git for version control. Show less