GenieWizard
An LLM-driven tool that simulates user commands to discover the features a multimodal app is missing, before developers build them.
Aug 2024 — CHI 2025I'm a co-founder and Head of Technology at Subtle Computing. I build voice-isolation, ASR, VAD models, etc. to help your device better understand you and respond only to you in extreme SNR scenarios. Our first product is Voicebuds: earbuds with near-perfect wearer speech isolation, for clear calls in loud environments and AI chat and dictation at a volume nobody around you can hear. At Subtle I lead machine learning, backend and infrastructure, firmware development, and automated hardware/software testing.
Before Subtle, I did my CS PhD at Stanford with James Landay and Monica Lam, building the frameworks that make multimodal interfaces (voice, touch, and GUI together) practical to develop: ReactGenie, GenieWizard, and AMMA.
An LLM-driven tool that simulates user commands to discover the features a multimodal app is missing, before developers build them.
Aug 2024 — CHI 2025A software architecture for AR guidance assistants that adapt to a user's progress, preferences, and capabilities via a learned user model.
Mar 2024 — IEEE VR 2024A programming framework that turns existing GUI state code into rich voice-plus-touch multimodal apps using LLM-generated parsers.
May 2023 — arXivFull-body tracking in virtual reality improves presence, allows interaction via body postures, and facilitates better...
Apr 2022 — CHI 2022DoThisHere accepts multimodal interaction to help with user’s cross-app tasks. The user can use voice...
Oct 2020 — UIST 2020Although state-of-the-art smart speakers can hear a user’s speech, unlike a human assistant these devices...
Apr 2020 — CHI 2020