About
I am a UC Berkeley student studying Data Science and Statistics, focused on building end-to-end systems: clean data pipelines, reliable infrastructure, and applied ML that’s evaluated and shipped like a product.
Current Focus:
- Applied ML (training signals, evaluation, feedback loops)
- Systems & data (pipelines, storage, reproducibility)
- Interactive tools (UIs and demos that make results usable)
My Engineering Philosophy
Complimenting my theorectical perspective on Data, my preferred approach in tackling any question or set of requirements is an analysis of pain points, what does the target customer prioritize then architecting and designing solutions whose features intend to resolve those pain points. In cases of optimization, I prioritize understanding frameworks and identifying bottlenecks, then modify existing systems/APIs and evaluate performance.
What I care about
Reproducibility - Clear “run it again” workflows, versioned data/artifacts, and minimal ambiguity.
Concrete Evaluation - Metrics, baselines, ablations, and the habit of asking “what would change my mind?”
Simple architectures that scale - Start with clean interfaces, keep components modular, and make it easy to extend without rewrites.
Polished outputs - Tools should be usable by someone else (docs, UI, sane defaults, good ergonomics).
How I work
- I design a thin MVP first (the smallest version that can be tested end-to-end).
- I prioritize instrumentation early (logs/metrics/tests) so iteration is fast.
- I write short technical notes as I go: assumptions, design choices, and next steps.
- I like projects where I can own the full loop: data → model/logic → interface → evaluation.
Interests right now
- Personalized training systems (difficulty control, tagging, feedback loops)
- Retrieval + ranking pipelines and evaluation harnesses
- Reliable automation for “messy” real-world workflows
- Game-like learning environments and progression systems