After three years building ML systems in industry, I wanted to go deeper. Production work taught me how to ship models, but I kept running into questions I couldn't answer about why they worked or failed. Graduate school was a chance to fill those gaps.
Choosing Santa Cruz
I picked UC Santa Cruz for its strength in AI and its proximity to the Bay Area research community. The program had the right balance: rigorous enough to build real expertise, flexible enough to pursue research that interested me.
The location mattered too. Santa Cruz is close enough to industry that you stay grounded in real problems, far enough that you can focus on longer-term thinking.
What I Studied
The coursework rebuilt my understanding of ML from first principles. Advanced machine learning, deep learning theory, computer vision, NLP. But unlike undergrad, I came with context. Learning about optimization algorithms meant more when I'd spent years watching gradient descent struggle on real data.
The distributed systems course was unexpectedly valuable. I'd built ML systems but never really understood the infrastructure underneath. Learning about consensus, replication, and failure modes gave me vocabulary for problems I'd encountered but couldn't articulate.
Research
My research focuses on medical imaging: making AI systems that doctors can actually trust. I work on CT scan denoising and model interpretability. The technical challenge is interesting, but what drives the work is the gap between what AI can do and what clinicians will accept.
A model that improves image quality means nothing if a radiologist can't understand why it made a particular enhancement. I've been exploring attention mechanisms and visualization techniques that make model decisions more transparent. The goal is building systems where improved performance comes with improved trust.
The Balance
Maintaining a 3.9 GPA while doing research taught me about focus. There's always more you could read, more experiments you could run. Learning to choose what matters, and accept that some things won't get done, was its own education.
The program ended in June 2025. I left with a deeper understanding of the field and clearer ideas about what problems are worth working on.