Tackling Fine-Grained Robot Manipulation Tasks Using Multi-Modal Reinforcement
2021 - present
I am working in the IRIS lab at Stanford University on a project that uses sound, image, and proprioceptive data to get robots to do relatively tricky tasks, including extracting a mug from a bag and grabbing a block from behind a wall. Paper and code coming soon.
Example of extraction
Monitoring Human Activity with Wi-Fi Metadata
2019 - 2020
I use machine-learning and CSI (Channel State Information) metadata from an ESP32 WiFi prototyping board to sense human movement in an enclosed space. Rough source code is available.
A sample of CSI data
Improving LSTM Neural Networks for Better Short-Term Wind Power Predictions
2018 - 2019
I created modified LSTM (Long Short Term Memory) Neural Networks and used auxiliary weather forecast data to improve ultra-short wind farm output predictions, for use in a smart grid. The paper is published in IEEE Renewable Energy and Power Engineering (REPE) Conference, 2019. Poster and source code also available.
Plot of prediction (blue) vs. real power (orange)
Interpreting Sign Language Using CNN-based Models
Example of a pre-processed frame
Non-Invasive Detection of Sudden Infant Death Syndrome (SIDS) Through Recurrent Neural Networks
2017 - 2018
I created auditory monitoring hardware and recurrent neural networks to track and analyze the breathing patterns of a sleeping person for signs of respiratory irregularity. Research poster, paper, and source code are available.
Monitoring hardware prototype
Intellistove: An Intelligent Stove that Prevents Cooking Fires and Exploding Glass Flasks in the Lab
2016 - 2017
I designed, fabricated, and tested a smart stove system that turns off when a pot boils dry. It is controllable via secure text message. The research poster, source code, and video demo are available.