Back to Home Research & Work

Blog & Insights

Notes from building real-time accessibility systems, evaluating large language models, and the engineering lessons along the way.

Federated Learning · Research

The "Slowest Client" Problem: How Adaptive Compression is Rescuing Decentralized AI

Why static compression is holding back Federated Learning—and how adaptive distortion-aware protocols achieve up to 89% data reduction while preserving model accuracy.

Federated Learning Lossy Compression Distributed Systems Green AI

Tutorial

A Beginner's Guide to Error-Bounded Lossy Compressors

Understanding SZ2, ZFP, TTHRESH, and FPZIP—the compression algorithms powering efficient scientific computing and distributed machine learning.

Tutorial Compression Scientific Computing

Coming Soon

Building ASPIRE: Lessons from Describing Slides in Real Time at a Live Conference

What it took to go from a Wizard-of-Oz study to a system that keeps blind and low-vision listeners in sync with a live speaker: slide detection, time reclaim, and what twelve participants taught us.

Accessibility Real-Time Systems HCI