Selected work

At that Google hackathon I ran straight into how hard Smart Code Writer actually was. The version I wanted would not just synthesize programs, it would correct its own mistakes while doing so. I took the problem to one of my professors, and he told me that what I was describing, at a high level, was the field of artificial intelligence.

So I went and studied it. I found that his own thesis work was in computer vision, applying state-of-the-art machine learning and AI, and I wanted to learn it properly. He said he would teach a course on it if I gathered enough signatures on a petition. I got the signatures, and he taught the course.

Ever since 2015, when I learned to code and finished my internship at Google, I have kept turning over the same question: what sufficiently intelligent, self-correcting systems mean for a single person, and how they change what one individual can take on when the work lives in a computational space. That question became one of the foundations of my PhD application, and it is the same one agentic development is now answering in practice.