- Problem
- AI tools can do the thinking for kids.
- What I built
- A maker studio. Kids choose, build, test, and explain a project. The AI coaches. It does not build for them.
- Evidence
- No learner results published yet.
- Still unproven
- Whether learners can show the skill without the AI.
Founder, SayMake · ex‑Apple, Microsoft
I build AI products that help people learn, create, and grow more confident.
Reach out for advising, collaboration, speaking, or build opportunities.
A few things I’m buildingThoughtfully made. Always learning.
From curiosity to something useful
Selected work.
SayMake
A place to become a builder.
Private preview- Open Making and Building StandardIdeas take shape.
Practice leaves evidence.Concept illustration · Draft standardOMBS
What does it mean to build well?
Experiment - ReviewFlowA person stays
in the loop.Concept diagram · Local experimentReviewFlow
AI suggestions. Human decisions.
Experiment 
SPEAR
Trust, checked at the boundary.
Experiment
Cue
An assistant for the everyday.
Experiment
- 3 Apple Maps launches
- My teams shipped Transit cards and fares, EV routing, and 3D Maps.
- 2 products from zero
- Co-led Bing Travel at Microsoft. Founded and built SayMake.
- 1B+ users
- Across Apple and Microsoft products.
How I work
I start with the people and the problem. I listen, explore what could help, and build a first version to test. My goal is to turn a promising idea into something useful, then learn from how people use it.
Lead problem
Skills people can prove
“Can this experience help someone do something valuable they couldn't do before, and can they demonstrate it independently?”
Second problem
Agents people can trust
“Can someone trust the result, inspect the evidence, correct mistakes and stay in control?”
Cue, ReviewFlow, and ResumeStellar: tested locally, not yet verified in another person's account.
- Problem
- Model proposals are suggestions, not ground truth.
- What I built
- A local review app. A model proposes labels. A person reviews and corrects them. Exports keep each decision, reason, and source.
- Evidence
- The public repo has setup steps and verification results. Demo data is simulated.
- Still unproven
- No real human benchmark exists yet.
- Problem
- AI app security is often asserted, not inspected.
- What I built
- Open-source LLM security middleware with a red-team evaluation harness.
- Evidence
- The public repo shows the source, README, and evaluation method.
- Still unproven
- Live models, new attacks, and real integrations. The checks do not cover output leakage or latency.
Cue
Status: Experiment
Not verified for this refresh- Problem
- Handing parts of my own daily work to an AI assistant.
- What I built
- A personal AI assistant built around my own daily workflow.
- Evidence
- Local tests in my own workflow.
- Still unproven
- Use outside my own workflow.
ResumeStellar
Status: Experiment
Not verified for this refresh- Problem
- Tailoring a resume to each job post takes time.
- What I built
- An experiment that tailors a resume to a job post.
- Evidence
- Local tests only.
- Still unproven
- Use in another person's account.
Other work
- Puddle. Practical next steps for everyday parenting questions.
What I'm looking for
I want to own a consequential AI product, end to end, inside a strong team.
End to end
- Customer problem
- Technical choices
- Launch
- Measured results
The mandate
- Direct access to customers
- Authority to make product and technical decisions
- Capable colleagues
- One outcome I am accountable for
My evidence: my teams shipped Apple Maps features. As a founder, I build products end to end.
Career
At Apple and Microsoft, I worked on complex systems. As a founder, I bring that experience to hands-on product building: connecting the vision to the technical details, bringing people together and taking ownership when the path is unclear.
SayMake
Founder & solo builder
2025 — now
Apple Maps
Senior engineering manager
My teams shipped iOS 16 Transit cards and fares, EV routing, and 3D Maps.
2018 — 2025
Salesforce
Senior software engineer
Built production fraud detection with deep learning.
2015 — 2018
Microsoft (Bing, MSN)
Senior engineering manager
Co-led Bing Travel from zero.
2007 — 2015
Written Thoughts
Notes on agency, learning, and trustworthy AI.
Get in touch
Reach out for advising, collaboration, speaking, or build opportunities.