Research

I am a PhD student in Eric Brewer's lab at the University of California, Berkeley.

Stasis

I am developing an extendible write ahead logging system. It is geared toward high-performance applications that require control over the on-disk layout of their data, or simply are a poor fit for existing storage mechanisms. Unlike past implementations, the library provides fine-grained, layered access to write-ahead-logging components, allowing applications to customize recovery semantics, in-memory layout, logging policies, and eventually distributed storage algorithms. The core interface closely reflects the underlying storage mechanisms, avoiding a number of bottlenecks typical of current systems.

I am also interested in using language based techniques such as static analysis, and application-specific optimization strategies to increase the overall reliability and performance of the system.

Resume

[pdf]

Class projects

Here are some of the course projects that I've worked on at Berkeley.

Web Traffic Analysis using Latent Dirichlet Allocation of Markov Chains

I applied this paper (longer draft copy) to the KDDCUP 2000 gazelle.com dataset, and proposed and tested an extension of the original model. The results were mixed, but the extension to the model may be interesting in other contexts, as the extended model performed well under certain constraints.

Aimless Documents

Egon Pasztor and I developed a scheme that automatically generates 'glue code' to allow existing software modules to interoperate. It includes a simple query engine and expert system, and an XML-based language that allows end-users and third-party developers to describe interfaces provided by existing software components. A simple bean shell runtime allows the system to be used interactively.

Encryption keys

I now prefer S/MIME, and have a thawte freemail certificate.

Here is my old gpg key: [gpg]