A fully automatic system that utilizes deep learning techniques to identify and record high-quality cellular images through microscope. The machine’s ability to conduct long-term experiments while maintaining consistent efficiency ensures the reliability of experimental results.
Production‑grade activity recognition for wearables using a cascade of lightweight temporal models and robust signal features. Data curation, augmentation, and Monte Carlo uncertainty provide accuracy, and stable user experience under real‑world noise. Engineered for low power, small memory, and low latency, the system supports daily activities, swimming, and gestures—deployed to 1M+ users for personalized health tracking and engagement.