I'm a Robotics Ph.D. candidate in the Chestek Lab at University of Michigan, working on brain-machine interfaces for finger movement decoding. I did my undergrad in Chile, majoring in robotics and computer science, and love working on projects that can improve people's lives through robotics.
- A deep-learning-augmented Kalman filter for high-performance intracortical decoding: we're studying KalmanNet, an algorithm that enhances the Kalman filter by adding deep learning to learn the measure of trust between physics and neural activity. We are seeing an improved performance in real-time finger movement decoding in primates, which is achieved by allowing the brain to have a lazy computation and only care about change.
- Compression or noise reduction: the role of dimensionality reduction methods in understanding the brain-muscle relationship: dimensionality reduction methods are widely used to study brain and muscle latent factors, but it is unclear whether these techniques are able to remove noise or are just an efficient way of compressing the data. We have found that, when trying to predict muscle activity from neural activity, dim. reduction methods were able to compress very effectively, but didn't show signs of actually denoising the data.
- Reliability and Minimal Detectable Change of Stiffness and Other Mechanical Properties of the Ankle Joint in Standing and Walking (2023, Gait & Posture, first author, available here)
- Balancing Memorization and Generalization in RNNs for High Performance Brain-Machine Interfaces (2023, NeurIPS, third author, available here)
- Error detection and correction in intracortical brainβmachine interfaces controlling two finger groups (2023, Journal of Neural Engineering, fifth author, available here)
- Breaking the barriers to designing online experiments: A novel open-source platform for supporting procedural skill learning experiments (2023, Computers in Biology and Medicine, first author, available here)
- Motor slacking during resisted treadmill walking: Can visual feedback of kinematics reduce this behavior? (2021, Gait & Posture, second author, available here)
- High Five: we created an open-source myoelectric upper-limb prosthesis fully made in Chile. The CAD and PCB designs are available online, as is all the code necessary to replicate the project. Look here and here for some awards we received for this project.
- Workshop on neurotechnology at UCSD, 2023: developed workshop to teach neurotechnology tools, such as feature extraction from ephys recordings and decoding models. We developed multiple Jupyter notebooks to walk the students through the most important concepts.
- Languages: Python, C, C++, Matlab
- Tools: ROS, LCM, Git, Linux, real-time systems