I am currently working as an Artificial Intelligence Researcher at UC Berkeley (advised by Professor Claire Tomlin). Previously I was a masters student in EECS at Berkeley (advised by Professor Jitendra Malik).
My research interests lie in artificial intelligence with applications in machine learning, computer vision, robotics, and climate.
Projects
Teaching
I assisted Professor Ronald Fearing in helping students design and build autonomous 1/10th scale line following RC cars. Students are challenged to design all circuitry (motor drive, power, custom pcb, etc.) , custom firmware, and algorithms (perception and control) in this semester-long, team project.
Miscellaneous
Realtime Convolution Algorithms
In this project I wanted to build a realtime digital convolutional reverb that I could use when I play guitar. I hope to build a "reverb pedal" by deploying ideas explored in this project to an embedded processor.
I found this paper that describes an efficient method for realtime convolution on single threaded operating systems. The method described cleverly uses Overlap Add Convolution with Decimation in Frequency Fourier Transforms and well chosen block sizes to dramatically reduce computation time, allowing for real time processing. I have implemented various convolution algorithms described in the paper in python.
The plot, which compares the time taken to convolve a fixed input of ~8 seconds with a fixed reverb signal, clearly demonstrates that the "finite delay line" method described is much more efficient than traditional the "uniform partitioned convolution" method for realtime signal processing.
Tube Guitar Amplifier
I spent a semester abroad studying at the Pontifical Catholic University of Chile in Santiago, Chile. While studying here I took a course entitled "Electronic Audio Workshop" in which I designed and built a vintage tube guitar amplifier. I based my design off the famous 1960's Fender Deluxe Reverb.
