Speaking
Voice AI, agent infrastructure, and the machinery I actually run
I present in the Portland and Vancouver AI community, and this is not new to me. Running my own consultancy through the 2000s I delivered over 40 trainings, online, in classrooms, and internationally in Italy, Canada, and Malta, essentially all of it on one open-source content management system. The online classes ran long and drew people from all over the world, including from the Department of Homeland Security, the United Nations, major universities, and various governments. The one I'd point at is Malta: the government there hired me, flew me in, and I spent five days with a room of their people. That one went well.
The thing that makes this work for me is that I don't use a script and don't want one. Being brought in to help a group figure something out is a comfort zone rather than a stretch, and the value is in reading what's actually happening in the room and adapting to what these particular people need. A fixed curriculum would take away the part worth hiring.
Every entry below links to the public listing that names me. If it isn't listed somewhere you can check, it isn't on this page.
If you run a group and want a talk on any of this, get in touch.
Talks and demos
Repurposing a programmable keypad for infrastructure and AI management
Cascadia Builders Club, August Demo Night, at North Bank Innovations
A live demo of the control surface I built on a cheap programmable keypad, and the multi-machine AI development ecosystem behind it: a Mac Studio at the center, a MacBook Pro, and an iPhone, joined over Tailscale so my agents can reach all of it. Also covered the Apple Watch app and the dashboard, and the hardware argument underneath it all, which is what this class of device has to be, what my first unit could not do, and what to buy instead. Everything running on that keypad is software I wrote. No vendor app is installed.
The Journey of Building Voice AI Learning
Portland AI Engineers, March Demo Night, at Multnomah County Library Albina
The story of UnaMentis from the outside in: why I started building it when the voice interfaces I relied on stopped serving the kind of long, hands-free learning conversation I wanted, what it took to get there, and what I got wrong along the way.
UnaMentis: On-Device Voice Models
AI Tinkerers Portland, Building Voice Agents and Conversational AI Stacks
Background on UnaMentis and the case for pushing voice models onto the device: what it takes to get latency low enough that a spoken learning session actually feels like a conversation, and what you give up and gain by running Pocket TTS and GLM-ASR locally rather than calling out to a provider.
This is not everything I've presented, it's everything I can point you at. More gets added as I track down the listings, along with slides and recordings where they exist.