Kevin Chen

Building shared frontend infrastructure and design system components for Patreon's creator-facing platform.

• Built a centralized, reusable Date Picker component for Patreon's internal design system, enabling consistent, accessible date selection experiences while reducing duplicated implementation across product teams
• Partnered with Studio engineers, product designers, and cross-functional stakeholders to define requirements, validate edge cases, and deliver a flexible component supporting diverse creator-facing workflows
• Contributed to shared frontend infrastructure by developing scalable UI components that improved design consistency, enhanced developer experience, and accelerated engineering velocity across multiple teams

Researched and implemented AI model compression techniques, bridging ML research and production inference infrastructure.

• Researched and evaluated AI model compression techniques, including fine-tuning and post-training optimization methods, to reduce model size, GPU memory usage, inference latency, and compute cost
• Built automated evaluation and benchmarking pipelines to compare compressed and baseline models, measuring accuracy, latency, throughput, memory footprint, and deployment readiness across realistic inference workloads
• Bridged machine learning research and production engineering by validating compressed models with LLM serving infrastructure (e.g., vLLM), achieving up to 2.5x compression while maintaining ~90%+ of baseline accuracy