Principal data scientist working on a variety of scientific and engineering problems related to biomedical engineering, robotics, predictive maintenance, and process control. Bringing machine learning skills to power engineering solutions.

Scientific program manager across several clinical neural engineering-related projects, ranging from human brain-computer interfaces to spinal cord and bladder stimulation inside the Rehab Neural Engineering Labs.

Working to improve brain-computer interfaces (BCI) for individuals with tetraplegia by studying the interface between BCI decoders and artificial somatosensory feedback via intracortical microstimulation.

Researched neural processing in mouse visual cortex using in-vivo two-photon fluorescent calciumimaging to record the dynamic activity of 200+ neurons per imaging plane. Recorded hundreds of repetitions of visual stimuli in order to capture the neural variability inherent in visual processing to better inform machine learning models. Additionally, used large field of view (1mm by 1mm) two-photon calium imaging to create maps of the cortical location that processes a particular part of visual space.

NanoVision Diagnostics is working to commercialize a novel electro-optical cancer diagnostic system that is able to detect molecular changes in a standard biopsy slide. I am working on as part of a team to integrate NanoVision Diagnostics' cancer detecting algorithms into their next generation data collection system.

Part One: Making brain-computer interfaces. How do you use 100 streams of neural information to move a cursor on a screen? We have to reduce the dimensionality of the data to remove noise and understand how the neurons work together to coordinate movement. We then check our ideas by modeling the time-varying activity of a large population of neurons.Part Two: How do you control a brain-computer interface? Right now, in order for a person or monkey to control a cursor with their neural activity, they need to maintain visual contact with the cursor. However, visual information is not helpful when you want to feel what you're doing - like grasping a glass of water. We are investigating nonvisual ways to provide information, including stimulating sensory areas of the brain or providing vibrotactile feedback on the skin.

We modeled the electrical activity generated by the hand muscles from stroke survivors in order to classify their intended grasp. Ultimately, an assistive orthotic could be made that would infer the person's intended grasp and then help their hand conform to that posture.