Principal AI leadership, embedded in your team

I join your team as the principal who carries the hard AI decisions and ships them.

I work inside your codebase, data, and constraints, from a blank page to a system under scaling pressure. I work with executives, product, and engineering to turn a mission into clear actions and move it forward.

Book an Intro Call

Principal applied ML work across Roblox (production ML, trust and safety, inference serving), Google AI (search and recommendation models for verticals like podcasts and recipes), and more. See the work

Ways to bring me in

Three ways to bring me in, from a single focused read to an ongoing seat on your team.

Focused reviews and recovery sprints are the best current entry points; ongoing fractional work is capacity-limited.

Ongoing fractional engagement

Fractional AI Lead

Currently at capacity for ongoing engagements

I lead your AI work as the product and system evolve, week over week, decision over decision. You do not need an existing engineering team or a finished system; I can help shape the mission and then build toward it. Fractional describes my flexibility about hours, not the depth I bring or the responsibility I hold.

New ongoing fractional engagements are currently at capacity. If this is the shape of help you need, send a note with the mission, what is blocked, and timing; I can tell you whether a future start or referral makes sense.

Send a note

Time-boxed, usually about 6 weeks

AI Initiative Recovery Sprint

I bring concentrated principal attention to an initiative that has stalled or lost its shape. I work inside the code, data, and constraints, separate the core delivery problem from the noise, and reshape the highest-leverage part alongside your engineers. A sprint can grow into the ongoing Fractional AI Lead role when the fit is right.

Focused review

Production Readiness Review

I help you figure out whether AI is the right bet for this product, and whether it solves a painkiller problem or just a vitamin, something everyone says they'd want but few actually need. I ask the sharp discovery questions to find that out before you commit the budget. Once it's worth doing, I get hands-on with the architecture, data, and evaluation that make it work, and I keep the cost of running it sane so it pays off.

Recent engagements

A sense of the work

Mental-health AI · embedded principal

Safety architecture for a conversational AI used by millions

Embedded as principal safety architect for a mental-health AI companion: decoupled safety, in-context classifiers, a pass/steer/block gate, and an append-only audit trail validated in shadow mode.

Read the case study

Disarray · fractional principal engineer

Agent memory and evaluation for autonomous ML

Worked with the team on all surfaces of agentic production ML, with more focus on two: the agent's memory and context persistence, and an evaluation discipline that separates genuine capability from reasoning, retrieval, and workflow failure.

See the full body of work

The first call

A short call to see if I can help.

Intro call, no slides. I want to understand the mission, where AI sits inside it, and whether I would be empowered to carry the decisions that move it. If we are not a fit, I will say so on the call.

Book an Intro Call

1 What is the mission, and is the role of AI in it clear yet, or still taking shape?
2 Are you starting from a blank page, or scaling something that already exists?
3 Who owns this work today, and what has it run into?
4 Would I be empowered to make and carry the technical decisions? Have access to a budget?
5 What does success look like, and how will we know when we are there?

Where I'm not the right fit

A few situations I'm honest about

I would rather be honest early than take work I cannot move. It's probably not a fit when:

There is no buy-in on the mission, or we cannot get aligned on what it is.
I'd be kept at arm's length from the people doing the work, expected to advise from outside rather than sit with the stakeholders.
The need is hourly staff augmentation with no outcome attached to it.
The work never connects to anything that ships.

Bring the AI work where the next decision has consequences.

If you have an AI initiative that needs a principal to carry it, the fastest way to know is to talk. Bring the messy version; that is the part I am here for.

Book the call