One Big Idea
My research publications are all in some way about changing assumptions and the good things that can happen when we do.
What do I assume that I haven’t tested? How do I benefit? Is it worth it?
Since you are here, think about that a little bit today.
Evidence is important
In statistics and machine learning, mathematical optimization, and high-performance algorithm design we (humanity) have gained a lot by leveraging strong assumptions about the problems we are working on. So great have been these gains that many assumptions are now baked into our everyday approaches. Often, we don’t stop to check and see if these assumptions are born out in evidence.
(10/26/2019) I was interviewed about whether AI will take away jobs from humans (spoiler alert: yep). My message: be flexible and remember that many AI trends are driven by demand. We all have responsibility for modulating our desires and taking a long term view (link).
(05/02/2019) I was interviewed for Leidos Q&AI Blog Series about some of my work using machine learning and artificial intelligence to design computer systems faster (link).
About My Education
I have been educated several times, and eagerly anticipate several more.
My PhD dissertation, Statistical Metrics of Hardware Security, is about better ways to tell whether computer chips are doing things correctly and privately.
My PhD in Computer Science was advised by Ryan Kastner at UC San Diego. I did science on computers. First I made guesses, then took measurements, checked those measurements with new statistics and tools that I invented, and published and presented the results.
I currently work as a senior hardware security engineer at Tortuga Logic in San Jose, where I help different companies who make computer chips be sure that their products do not have security problems.
I used work as a senior research scientist at Leidos in San Diego, where I solved problems with national impact to help people live happier lives. At Leidos I led a successful project: we made a program that automatically transforms software that can only run on one processor into software that can run on many different kinds of processors at the same time (called heterogeneous systems). Many kinds of software run much faster on heterogeneous systems, but deciding how to write the software is hard. My project made it much easier. I also made contributions to algorithms for counter-adversarial machine learning, automated experimental design, and real-time machine learning.
I have previously been employed as a software developer and audio engineer.