Projects
Triaging TB Patients using Image Classification.
Healthcare
-
Visualised the dataset comprising of about 662 chest x-rays using Zeppelin Spark.
-
Performed histogram equalization and multiple data augmentations to enhance the dataset.
-
Created a deep learning model using the VGG16 architecture and the TensorFlow library running on top of AWS EC2 instance. It gave us a loss of 2.76 and an accuracy of 82.5% during validation.
Predicting CVD for Osteoarthritis Patients.
Healthcare
-
The dataset comprised of about 400,000 survey response records and 600-1200 features. Cross checked literature-derived features with Lasso-suggested features to select 19 most significant of them.
-
Resampled the dataset to balance positive & negative cases, and applied sample weights to scale for population.
-
The XGBoost model was 74.1% sensitive at an accuracy of 74% for patients between ages 50 to 64 and annual income less than $15,000. They’re three times more likely to get CVD.
Process Optimization.
Supply Chain
-
Gathered the client Omega Neutraceuticals' data through different sources and interviewed management to determine the problem statement.
-
Designed a self-learning algorithm to better forecast inventory, and a performance-based payroll management and reward system that tracks productivity.
-
The system should reduce inventory prediction error rate by 15% and increase productivity by 25%.
Alumni Database System.
Academia
-
Followed Agile development model using constant feedback from faculty and a random sample of students.
-
Analyzed all business functions and created a CRUD functionality database by scraping for details on LinkedIn.
-
Received the grade A for the course Database Management Systems.
