Projects
A selection of work spanning machine learning, GPU programming, and systems.
Machine Learning
Low-level GPU API for mapping machine learning workloads to GPU clusters.
Tensor decompositions in Torch. Applied to TensorFaces paper implementation.
Implementation of different metric learning algorithms with GPU-accelerated comparisons.
Classification of images into different scene categories.
Using SVMs with MFCC and DWT features to classify singing voice vs. instrumental sounds.
Scale-space classification for image segmentation. Cloud Motion Vector generation for prediction.
General
Data structures + algorithms = programs. A collection of fundamental implementations.
Lossless compression pipeline featuring LZW data compression and wavelet-based image compression.
Systems & Architecture
Platform simulator for architectural design space exploration of heterogeneous systems.
Trace-driven simulation of stream buffers for data cache prefetching.
Full compiler developed in Standard ML for the Tiger programming language.
Implementation of range compression algorithm on GPU for SAR processing.
Evaluation and comparison of different sorting algorithms for performance.
Wavelet transform in integer domain for compression on fixed-point BF-535 processor.