Zachary Englhardt

PhD Student @uwcse | AI, Health, HCI

[Zachary Englhardt @zenglhardt1]

@vysrini Models + dataset on @huggingface, code on GitHub:
💻 https://t.co/kThkSUpqUE
🤗 https://t.co/K1PtuG30hu

Also with @VikramIyerUW and Shwetak Patel

[Zachary Englhardt @zenglhardt1]

What if your real-time voice agent could reason like a frontier model?

Conversational Infill lets a tiny on-device model talk while a frontier model thinks in the background: fast and responsive, grounded in frontier reasoning.

w/ @vysrini
📄 https://t.co/anfJCk6vrB
🧵👇

[Zachary Englhardt @zenglhardt1]

@vysrini The Talker is fast too: millisecond time-to-first-response, up to 19x quicker than the Reasoner while also hiding RAG and tool call latency.

In our live study (n=18, Apple M2), users rated it on par with frontier models and found it far more responsive on harder tasks. 🧵

[Zachary Englhardt @zenglhardt1]

@SpotifyCares ik this has been brought up before, but after 5 years of subscribing to premium, I hit the 10,000 song limit last month. Apple Music allows 100,000 songs, so I’m wondering if there is any plan to update Spotify. If not I think I will switch.