- $\cdot{}$ Avaak Inc. cameras that communicate wirelessly to a base station.
- The base station also communicated back to central servers. Under heavy load,
- communication would break down. They needed a way to generate such loads and
- log the communications and asked me to simulate the cameras and the base
- station. I set out simulating the camera first since it only communicated
- with a base station and I could have a real one at my desk during
- development. Also once it was developed it would make developing the base
- station easier. All of this went according to plan, but communication between
- the base station and the central server was much more difficult than
- anticipated. The central server had bad logging facilities and it was not
- feasible to modify it. To work around this, I developed a server that could
- run locally to facilitate the base station's server communication development.
- The simulators worked well and logged in a very readable format all
- camera, base station and server communications. Hundreds of simulated
- cameras revealed that the central server's memory was maxed out with a high
- number of connections.
- \newline{}
- $\cdot{}$ After using the simulator to diagnose problems like the central
- server's memory maxing out, Avaak wanted to use the tool to monitor the
- health of the system. This was a relatively easy addition that just needed
- simulate a round trip of a picture taken on a camera and then being visible
- on the website. This system emails logs of minor problems and sends a text
- in addition to a more detailed email for major problem. On catastrophic
- failures, I included myself on the emails. Several months after I left I
- received emails from this tool, so I knew it was still doing its job. At the
- end of my work on this project, it weighed in at 3kloc of ruby.
- \newline{}
- $\cdot{}$ Mike Sailor's chem lab at UCSD developed reusable chemical sensors.
- They enlisted the help of Avaak to figure characterize the sensor's changes
- when an agent was introduced to the environment. Working with Serge Belongie
- and Clark Guest, computer vision experts at UCSD, I developed C and matlab
- classifiers. These classifiers generated data which we analyzed using
- invariants, and later PCA, and Fischerfaces. The conclusion of this analyses
- was used to justify further research by Mike Sailor and his lab into their
- reusable chemical sensors.
+ \begin{itemize}
+ \item
+ Avaak needed to stress and log its wireless camera and server to pin down the
+ communication breakdown. My tools revealed that the server's memory maxed
+ under a high number of camera connections.
+ \item
+ Sometimes the servers would be up but not doing their job correctly, so Avaak
+ asked me to modify the simulators to monitor the servers more intelligently.
+ Several months after I left, it would still send me detailed failure reports.
+ \item
+ A chem lab at UCSD wanted to characterize a reusable chemical sensor's
+ appearance change. Working under the direction of several professors, I
+ developed quantifiers for the amount and type of change of the sensors.
+ \end{itemize}