The KapLabs
Passion Projects
Closing the Sensory Void - Reducing Prosthetic Rejection rates through innovative adaptive spiking feedback
Problem #1
Current prosthesis design and lack of somatosensory feedback prevents prostheses embodiment leading to high rejection rates (up to 44%) and frustrates amputees.
Closing the sensory loop is one of the hardest unsolved problems in prosthetics.
Problem #2
High cost and complexity of a prosthesis research platform prevents democratization of technological research needed to improve these systems.
The Key Innovation
-
A Bidirectional, closed-loop neural interface for a low cost prosthesis with multimodal sensing.
-
This novel engineering solution not only learn gestures and adapts to the user intent in real time (<60s), it simultaneously processes environmental data (temperature and pressure) to provide real-time and adaptive haptic feedback.
-
Drastically reduces the learning curve and minimizes one of the leading cause of device abandonment by closing the sensory loop.
-
-
Low cost research platform: Realizes an otherwise complex and expensive platform in a low cost budget (~150 USD) making it accessible to masses for continued development and research
Congenital Heart Defects - Diagnosis with Machine learning and AI
Congenital Heart Defects (CHDs) are a significant health challenge, contributing to the highest rates of infant mortality among birth defects globally. It is estimated that approximately 1.35 million infants are born with CHDs each year, with around 25% of these cases being classified as critical heart defects. A primary contributing factor to mortality in these cases is the undetected nature of CHDs during prenatal evaluations, which can lead to serious health complications shortly after birth.
Leveraging echocardiogram video datasets from Stanford’s AIMI Center (EchoNet-Pediatric) and the Fetal Heart Academy, I engineered a Convolutional Neural Network (CNN) utilizing a DenseNet-264 architecture. The model classifies fetal echocardiograms to determine the presence of a CHD and identify specific conditions like Ventricular Septal Defects (VSD) with high diagnostic accuracy. To bridge the gap between data science and clinical utility, I designed a prototype user interface featuring an LLM-enabled clinical chatbot to provide physicians with immediate treatment pathways and recommendations. This project, which won 4th Place in Innovation across Oregon, represents my commitment to combining advanced machine learning with clinical cardiology to protect newborns.
Generative AI and Healthcare Clinical Desicions
Generative AI and healthcare Use cases!
