Led by Surya Guthikonda, Santhoshini Pranathi and Imrankhan. Part of the Cohere Labs Open Science initiative https://cohere.com/research/open-science
Hosted By
Building a 24/7 Multimodal Personal Agent, Part 3 of 3: Voice, Bolt-Ons and What Comes Next
Registration
About Event
This is the final session of a three-part coding series on building a personal agent that listens, sees, remembers and acts, entirely on-device with open models like Gemma 4 running through llama.cpp. Building on Parts 1 and 2, this session gives the agent a voice and makes it extensible.
We cover:
Speech output, interruption handling and a wake word
Proactive notifications
A native skill interface for adding new capabilities
MCP connectors: paste a standard config, and see where a small model starts to struggle
Each component comes with its own evaluation. Live tasks: spoken proactive alerts, and a voice-set reminder created through MCP. We close with the roadmap and how this compares to pi, OpenClaw and Pi Durable.
Led by Surya Guthikonda, Santhoshini Pranathi and Imrankhan. Part of the Cohere Labs Open Science initiative https://cohere.com/research/open-science
Hosted By