I turn "AI could probably do this" into things that actually [work ]

Services

How I can help (without breaking things)

I've shipped enough AI features to know what works. Let me save you some headaches.

AI Product Strategy

Helping you figure out if AI is the answer or just a shiny distraction. Spoiler: sometimes it's both.

Team Enablement

Teaching your team to fish (with prompts). Workshops, reviews, and ongoing 'why is it doing that?' support.

Integration Support

Bolting AI onto your existing product without breaking everything. Error handling included, sanity optional.

What I Actually Do

Take your AI dreams, ground them in reality, and ship something people will actually use. No magic, just work.

Context Engineering

The secret sauce between 'cool demo' and 'useful product.' Turns out, what you feed the AI matters a lot.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

Start Manual, Then Automate

Humans do it first, robots do it later. I've automated myself out of enough problems to know better.

Context Is Everything

Garbage in, garbage out. I'm borderline obsessive about what the model sees and when it sees it.

Design for Failure

AI will do something dumb eventually. I build the safety nets before you need them. You're welcome.

Iterate With Real Users

Your customers are smarter than your assumptions. Ship early, listen often, adjust weekly.

Measure What Matters

Token counts are for nerds. Did it save time? Did users love it? Did it actually solve the problem?

Document Everything

Future you will thank me. So will your team when they inherit something they can actually maintain.

Keep Humans in Control

AI is the copilot, not the pilot. Every system I build has a human somewhere making the real calls.

Ship Fast, Learn Faster

Perfect is the enemy of 'in production.' I'd rather ship something good than polish something forever.

Build for Change

Today's GPT-4 is tomorrow's GPT-ancient. I design systems that can swap models without a rewrite.

Communicate Clearly

I speak fluent engineer and fluent human. Everyone leaves the meeting knowing what's actually happening.

How I Work

Hard-won lessons from projects that went sideways (and the ones that didn't).

FAQ

Stuff people usually ask

The questions I get most often, answered before you have to awkwardly bring them up on a call.

Latest posts

Brain dumps and lessons learned

Things I figured out the hard way, written down so I don't forget (and maybe you won't have to learn them the hard way too).

Get in touch

Got an AI problem? Let's chat.

Tell me what you're building, what's not working, or what wild idea you can't stop thinking about. I'll get back to you within 24 hours (probably faster, I check email too much).