Contributed to a Unity retargeting project: designed data collection approaches using the Rokoko Suit system, implemented a UDP-triggered pipeline to synchronize motion and audio, developed retargeting software in Unity, and advised on motion capture experiment research approaches. Supported delivery of Android and web applications by building demos, integrating features, and collaborating with cross-functional teams to ensure smooth implementation and communication.

Built end-to-end AI experimentation pipelines including test plan design, data collection, and automation of preprocessing and evaluation, reducing experimental runtime from days to hours. Conducted comparative evaluations across model variants, investigated edge cases causing false or missed alerts, and summarized findings to guide iterative model selection.

Contributed production-quality Android and Flutter AI samples to TensorFlow Lite and MediaPipe open-source repositories, collaborating with a Google Developer Advocate to demonstrate on-device inference best practices. Focused on applied research of AI on Android devices, shared findings through technical sessions and cross-team knowledge sharing. Led implementation of an online radio feature with real-time audio streaming (Agora SDK) and live chat (Firebase Realtime Database).

Led development of a GPS-based golf course mapping module, implementing real-time distance estimation and club recommendation logic and delivering a client demo with zero critical defects. Contributed to an internal Android app for banking telesales using Kotlin and MVVM, integrating backend services, Google Maps, media upload/download, push notifications, and resolving device-specific camera compatibility issues. Contributed to early-stage Android projects for Japanese enterprise and healthcare clients building API-driven features, custom UI components, data visualization, and media integrations.