Portfolio
For some of the projects in this portfolio, certain information has been withheld or not disclosed to protect the confidentiality of the respective organizations.
Machine Learning
-
Using Feature Selection to Predict MIC Values with Neural Networks
This study used XGBoost and Neural Network models to predict the minimum inhibitory concentration (MIC) values for 13 antimicrobial agents. The XGBoost model selected 29 important features, which were then used as input for the Neural Network model.
-
Automated Plate Reader
A pipeline that comprises several Machine Learning models and image-processing steps. The pipeline allows users to input images of plates and accurately predict wells with growth. The primary objective of this project was to enhance a lab's capabilities, reduce wasted panels, and improve testing efficiency.
-
Bacterial Antibiotic Adversarial Genetic Algorithm (BAAGA)
This study tried generating new bacterial DNA sequences and new antibiotic compounds. The goal was to see if future resistance can be simulated, and if the generation of new antibiotic compounds could also be simulated.
-
Predicting resistance using gene expression and mutations
The project had two objectives. Firstly, to ascertain whether the resistance of a bacterial isolate to various antibiotics could be predicted based on gene expression or mutations (SNPs) in a specific set of genes. Secondly, to investigate whether specific mutations present in the genome of a bacterial isolate could provide resistance to an antibiotic.
-
Isolate Outlier Detector
This pipeline was built using Machine Learning and statistical analysis to identify bacterial isolates that were outliers during antimicrobial testing. These outliers could be from testing errors, undiscovered resistance mutations, or failed tests. By detecting these outliers, the lab can reduce costs associated with unnecessary testing and identify potential new resistance mutations with minimal human intervention.
-
Identifying Beta-Lactam Resistance with Neural Networks
A research study aimed to anticipate whether a bacterial isolate would be classified as susceptible or resistant to a specific antibiotic. To ascertain the optimal algorithm for predicting susceptibility, both Decision Trees and Dense Neural Networks were employed. Additionally, correlation analysis was conducted on demographic information to identify any demographics that could serve as input.
-
Year, Location, and Species to Predict MICs with Beta-Lactamase Genes
This study aimed to evaluate whether dividing the input data based on species, year, and/or continent would enhance the accuracy of Random Forest or K-Nearest Neighbors in predicting antibiotic resistance (MIC values).
-
Parallelize Lasso with MPI
A study was conducted to distribute the Lasso algorithm, an extreme Machine Learning technique, using the Message Passing Interface (MPI) on a High-Performance Computer.
Pipelines
-
Annotation Pipeline
A pipeline was built to capture annotations from a bacterial isolate against a known reference gene. This pipeline then allowed different data visualizations to analyze trends easier.
-
Bioinformatics Pipelines
This website was designed to host numerous bioinformatics pipelines, providing Bioinformatics Scientists with fast and easy pipeline building, creation of new ones, and maintenance of existing ones. These pipelines operate on AWS, ensuring minimal downtime for lab personnel due to their large parallelism capabilities.
-
Automated Plate Reader
A pipeline that comprises several Machine Learning models and image-processing steps. The pipeline allows users to input images of plates and accurately predict wells with growth. The primary objective of this project was to enhance a lab's capabilities, reduce wasted panels, and improve testing efficiency.
-
Isolate Outlier Detector
This pipeline was built using Machine Learning and statistical analysis to identify bacterial isolates that were outliers during antimicrobial testing. These outliers could be from testing errors, undiscovered resistance mutations, or failed tests. By detecting these outliers, the lab can reduce costs associated with unnecessary testing and identify potential new resistance mutations with minimal human intervention.
-
Jenkins Auto-grader
For a Data Structures class, a pipeline was developed to automatically grade the Java homework of students. The students can work on their homework and periodically push it to a GitHub repository specifically assigned to them and the homework assignment. When the assignment is due, a Jenkins workflow will initiate automatically and grade all homework repositories.
Websites
-
West Nile Spread
This project was aimed at gathering comprehensive data on the spread of the West Nile virus, including the factors that influence its distribution and the impact it has on the surrounding region.
-
Avian Migration
Extensive migration data on numerous bird species was gathered to study migration patterns with climate data.
-
Health Co-Inquiry WebCrawler
Every day, this project scours a roster of websites, gathering user-generated text data for health research. The data is then analyzed in various ways to uncover insights into daily life with various illnesses.
-
Annotation Pipeline
A pipeline was built to capture annotations from a bacterial isolate against a known reference gene. This pipeline then allowed different data visualizations to analyze trends easier.
-
Bioinformatics Pipelines
This website was designed to host numerous bioinformatics pipelines, providing Bioinformatics Scientists with fast and easy pipeline building, creation of new ones, and maintenance of existing ones. These pipelines operate on AWS, ensuring minimal downtime for lab personnel due to their large parallelism capabilities.
-
Simple Budgeting
Effortlessly manage your finances and keep track of your progress, identifying areas where you excel and those that require improvement.