Builder. AI engineer. Founder of Sanford Labs.
The interesting part of AI is what you can build with it.
I work where models meet real systems. Built, tested, rebuilt. From the agent to the infrastructure underneath it.
An idea. A working system.
Something useful in someone else’s hands.
I understand an idea better when I can make it work.
That’s what keeps pulling me into AI. Memory, tools, context, permission: each opens up a new possibility. The interesting part is getting them to work together well enough to be useful.
I move between the product and the systems underneath it. Build a version. Put it under pressure. Find the part that needs another way of thinking.
I’m a CIO and the founder of Sanford Labs, with more than twenty years in engineering and technology leadership. Building is the thread through all of it.
Agents that remember. Conversations that connect. Information that becomes part of a room.
Different projects. The same curiosity about what useful intelligence can become.
Useful things happen
between disciplines.
What makes an agent useful over time?
Persistent AI bots with memory, scheduled routines and their own cloud computer. They ask before they act.
Explore JamBotsMemory, tools, execution and permission all have to work together. That is the engineering problem that makes agents interesting to me.
Where can AI make everyday work better?
Outlook, Gmail, IMAP and Teams in one inbox, with Claude suggesting where mail goes. Nothing moves until you confirm.
Explore JamCommsA practical place to explore assistance that gives people control over the decisions that matter.
What happens when software becomes part of a room?
Live wallboards for TVs and Raspberry Pi, with data and alerts pushed through scripts, a REST API or AI agents over MCP.
Explore JamBoardsSoftware becomes more interesting when its output leaves the chat window and becomes part of a room.
Built independently at Sanford Labs.
More from the labA new capability is exciting.
What we choose to do with it matters more.
A convincing response is a beginning. I care about what happens when the context is incomplete, a tool fails, or the model gets something wrong. The surrounding system has to make those failures visible and recoverable.
Capability and permission are different design decisions. Give an agent useful tools, clear boundaries and a way to ask. Human approval belongs where the consequences justify it.
Some problems need reasoning. Others need a query, a rule or a better interface. I want AI where it earns its place, and straightforward software everywhere else.
Prompts, retrieval, tool calls and infrastructure all shape the result. I want to follow a decision through the system, inspect the evidence and improve the part that actually broke.