San Francisco, California, United States
Working on large-scale multimodal video understanding infrastructure, focusing on both production indexing systems and emerging cognition capabilities. • Scaling multimodal video indexing infrastructure I contributed to the development of next-generation video indexing architecture designed to support multimodal foundation models in production. My work focused on improving system scalability, reliability, and cost efficiency through cloud-native deployment, autoscaling infrastructure, and distributed processing pipelines built on Kubernetes. • Building video cognition and agent systems More recently, I have been developing systems that enable reasoning across multiple videos by transforming multimodal outputs into structured knowledge representations. This includes building pipelines for knowledge graph construction and improving agent performance through retrieval and infrastructure optimization, enabling complex cross-video understanding workflows.
Joined Twelvelabs as an early engineering member during the pre-seed stage, helping build the company’s core video understanding platform from initial prototypes to scalable production systems. I worked on distributed video indexing pipelines that process decoding, multimodal model inference, and metadata generation at scale. My responsibilities included infrastructure development, backend services, and model runtime integration, enabling reliable deployment of multimodal AI capabilities. During the early phase, I also contributed to product development, including web applications for video manipulation and semantic search.
Worked as a machine learning engineer developing web-based information analysis systems for large-scale intelligence workflows. I built data ingestion and web crawling pipelines and developed backend services supporting real-time analysis and visualization. My work integrated machine learning models into operational systems to improve the searchability and analysis of multimedia data sources.