Current Research Directions

Dimensionality reduction of neural population data

(Yu Lab, CMU)

Today we are able to record neural activity from around a hundred neurons at a time. In order to analyze and interpret this high-dimensional data, we try to find a lower dimensional representation that can adequately capture the variability seen in the data. I am exploring different techniques that can extract these low-dimensional representations from neural spike train recordings.

Neural decoding

(Yu Lab, CMU)

Decoding is a critical step in the functioning of any neural prostheses, where decoders translate neural activity from the brain into control signals for guiding prosthetic devices, such as computer cursors and robotic limbs. I am interested in algorithms that can lead to improved performance of neural prosthetics.

Past Projects

Constraint Aware Robotic Cloth Folding

(Robot Learning Lab, UC Berkeley)

We developed a technique to enable robots to perform the tedious chore of folding clothes, using a cloth model that allows us to reason about the geometry rather than the physics of the cloth. Our work enables the robot to search for a time optimal solution with a richer set of primitives it can execute to complete user defined folds. Videos of a WillowGarage PR2 robot folding clothes using our technique, as well as a link to our ISER 2012 paper are available here.

Autonomous tennis ball bot

(Robot Learning Lab, UC Berkeley)

The aim was to build an intelligent, autonomous tennis ball-boy/girl created using low-cost, off-the-shelf equipment. I implemented search based motion planning algorithms for this project. A technical report on this project is available here.

CINEMA CubeSat

(Space Sciences Lab, UC Berkeley)

TRIO-CINEMA is a three spacecraft CubeSat mission to study Space Weather by making multipoint measurements of the magnetic field and energetic ions and electrons in near-Earth space and imaging ring current particles. I helped create preliminary Specs and Interface Control Documents for the STEIN particle detector. I also wrote some FPGA modules in VHDL for the alpha version of STEIN.

Multicore Gridding for MRI image reconstruction

(General Electric Global Research, John F Welch Technology Center, Bangalore, India)

In this project, we worked on a fast, multi-core CPU implementation of a 3D Gridding Algorithm that optimizes for speed without sacrificing accuracy.