Austin Cai

Spearheaded transformer-based models and real-time AI decisioning for SaaS spend management and support automation, reducing response time by 50%, increasing engagement 25%, and cutting support costs within six months.

Adapted CLIP to multilingual support and deployed scalable image retrieval pipelines, improving image search precision 22% and reducing inference latency by 30% via ONNX conversions.

Researched autonomous vehicle decision-making using POMDPs and RL, improving pedestrian detection accuracy 20% and reducing decision latency via Julia-based POMDP stacks.

Supported AI/ML initiatives by developing models, evaluating performance, and contributing insights that shaped the company's AI strategy.

Collaborated on image classification research, developing data augmentation techniques and experiments that improved model robustness and supported publications.

Built end-to-end pipelines for predicting Dasher sign-off behavior and demand quantiles, increasing driver retention 12% and improving delivery forecasting accuracy 20% during peak periods.

Migrated server codebase to Spring Boot, added Prometheus metrics, and implemented caching to reduce latency in conversation generation and improve scalability.

Developed tools to parse historical labor data for NASA mission timeline and cost evaluation, automating processes and saving ~10 maintenance hours per week.

Built applications for international staff to parse warehouse data, implemented admin access controls, and streamlined supply chain operations across global locations.

Completed a Bachelor of Arts in Symbolic Systems with a focus on artificial intelligence and related interdisciplinary coursework.

Completed a Master of Science in Computer Science with a specialization in computer systems.

Completed a Bachelor of Arts in Modern Chinese History focusing on historical and cultural studies of China.