Alex Dils
"And still 24 hours, maybe 60 good years, it is really not that long a stay."
I am a UC Berkeley computer science student. I work on computer vision and medical AI, with an emphasis on robustness,
bias, augmentation, and segmentation.
I have been a research intern with Stanford Medicine since 2022. I also build deployed ML systems through Appraise AI.
Projects
Multimodal pricing engine for resale marketplaces, built for scale and production monitoring.
Co-founded and led engineering: multimodal model + feature pipeline, high-throughput ingestion, and evaluation loops that reflect real marketplace shifts.
Designed for fast iteration while keeping reliability and latency constraints in view.
Astro ML sandbox: experiments, baselines, and reproducible training/eval runs.
A compact research-engineering repo for iterating quickly: clean data loaders, consistent evaluation, and small ablation-friendly experiments.
Designed to make comparisons easy and results replicable.
Static personal site: fast, accessible, and easy to update without a framework.
Lightweight HTML/CSS/JS with a theme switcher and external-link transitions.
Built to load quickly, work well on mobile, and keep content edits simple.
Speed reading practice tool with pacing controls and low-distraction UI.
A small web app for training reading speed and comprehension.
Includes adjustable WPM, short sessions for consistency, and a simple interface that stays out of the way.
Contact
dils@berkeley.edu