There are three depths you can bring me in at: decide, design, build.
You do not need to know which one you need. We can figure that out together.
1. AI Systems Advisory
Deciding whether it is worth building
You are considering an AI initiative, preparing to commit a budget, or reviewing a proposal from an external contractor.
I look at the problem, the proposed solution, the data and processes behind it, and the organization’s ability to use and maintain what gets built.
The key questions are often simple:
- Is the data good enough?
- Does AI actually add value here?
- What happens when the system is wrong?
- Who will own and operate it?
- Can the organization support it?
For outsourced work, I can also help define the scope, responsibilities, acceptance criteria, and what “done” means.
What I commit to
- A written assessment with the key risks and the path I would take in your position.
- A clear recommendation, rather than a menu of possibilities.
- An honest account of what is still unknown or missing.
- Technical specifications and acceptance criteria when an external team is involved.
- Advice that stands on its own, whether or not you hire me to build the system.
I do not advise on: your market, sales strategy, or product-market fit. Those are different problems.
2. AI Systems Design
Turning the decision into a buildable system
You have decided to build. Now the important question is: what should the system actually be?
I define the architecture, data requirements, component boundaries, technical approach, evaluation, and failure handling — including where not to use machine learning.
I also consider the system around the technology: workflows, ownership, maintenance, and how people will actually use the result.
What I commit to
- A design that covers architecture, data, components, failure handling, evaluation, and implementation sequence.
- Success criteria defined before the build starts.
- A design another development team can build from.
- A system sized for the organization that has to run it.
- A prototype where the most uncertain part needs to be tested.
- A clear recommendation when part of the solution does not need machine learning — even when that reduces my scope.
3. AI Systems Engineering
Building it into something usable
You need the system built, tested, and deployed.
I build the parts end to end: data pipelines, models, APIs, services, evaluation, containerization, deployment, and integration.
When the project needs an interface, I build that too. The model is not the product; the usable system is.
I then hand it over properly.
One project built and documented this way end to end: voice-assistant.
What I commit to
- Working software in your infrastructure, not just a prototype.
- Source code, tests, evaluation results, deployment configuration, and documentation as part of the delivery.
- A handover your team can actually use to operate and maintain the system.
- Clear boundaries around what I can and cannot responsibly take on.
- Containerized ML services, edge and embedded integrations, and end-to-end AI systems at the scale most organizations need — but not heavy multi-region distributed infrastructure.
- If I stop believing the project can succeed, I will say so, even after work has started.
Not sure which depth fits? That is what the review is for. Send me the problem.