AI Adoption & Production Engineering

SOFTWARE
THAT WORKS.

I help engineering teams get useful work out of AI in the codebase they already have, and I build the backends that have to keep running underneath it. Mostly fintech, e-commerce, and data-heavy systems.

Plenty of AI pilots look fine in a demo and go nowhere after that. I've spent 18 years on payment and billing systems, where mistakes cost money. You learn what can be handed to a model and what still needs ordinary code.

~6
Weeks
Typical time to first release
18+
Years
Writing production code
48hr
Response
Usually faster
Go • Java • Scala • Kotlin • TypeScript • PostgreSQL • Kafka • Kubernetes • AWS • GCP •

AI Work I Take On

I work in your codebase, with the team that owns it.

01
2 to 4 weeks

Set Up Coding Agents In Your Repo

Most teams have the licences and use them like autocomplete. I add what's missing: context files the agents read, CI gates that catch their mistakes, rules on what they can touch. Everything an agent writes still goes through your normal code review.

I leave it working, write down how to use it, and walk the team through it.
02
4 to 8 weeks

Build An AI Feature For Production

Retrieval, classification, summarization, or inference that runs on the device. Getting any of it past the demo stage takes a cost ceiling, a latency budget, evaluations in CI, a fallback for when the model is wrong, and a hard line around customer data. That's most of the work, and it's the part I build.

It ships with tests and dashboards, so you can see when it starts misbehaving.
03
1 to 2 weeks, fixed fee

AI Readiness Review

Short and scoped. I look at your systems and your team: where AI pays for itself, where it just burns money, and what to fix first. Usually the data needs fixing before the model does.

I write up what I find, rank it, and include cost estimates.

The first call

I don't sell strategy decks. If the problem needs a database query, I'll tell you to write the query, not buy a chatbot. If AI isn't your bottleneck, I'll say so. The call is free either way.

How I work

The demo
always works.

Anyone can get a good answer out of a model in an afternoon. The work is what happens when it's wrong, what it costs at ten times the traffic, and whether your team can still change the code six months later.

Frostwood Labs is a one-person studio. I built and ran a VoIP SaaS on my own, billing and on call included. That's where most of my opinions about production come from.

I read every line

Agents draft it, I review it. Nothing ships that I haven't read.

Production experience

I've been paged at 2 AM, debugged billing edge cases, and run the painful migrations. That's what I look for when I review model output.

Built to be maintained

Fast doesn't help if your team can't change the code later. I build what I'd be willing to maintain myself.

Projects

Products I built and ran myself.

CallerBooth

Browser-first VoIP built on Twilio, with smart voicemail and AI call transcription. I ran the whole thing solo, including billing and on call.

SaaS Shut down
SD

Screenshot Deck

iOS app that searches your screenshots by what's inside them. Vision OCR, CLIP embeddings, and semantic search running entirely on the device, with nothing sent to a server.

On-device AI Unreleased

What I Do

The engineering the AI work sits on top of. This is where I have the most experience.

01

Backend Systems

APIs, event-driven architectures, and data pipelines. I focus on systems that have to handle load and stay maintainable over time.

Backend API
02

Web Applications

Full-stack development when needed. Plain interfaces that don't get in the way.

Frontend Full-stack
03

Mobile & On-Device AI

Native iOS, plus AI that runs on the phone instead of a server. Usually the right answer when customer data can't leave the device, or when per-call API pricing doesn't work at your volume.

iOS CoreML
04

Data & Infrastructure

Cloud infrastructure, data pipelines, observability. The unglamorous work that keeps everything running.

DevOps Data
05

Technical Consulting

Architecture reviews, due diligence, helping teams get unstuck. Sometimes an outside perspective helps.

Consulting Architecture

Something else?

If your project doesn't fit these categories, ask anyway. I'll say quickly if I'm not the right person.

Contact Me

How Projects Work

01

Understanding

Week 1–2

We talk through what you're building, what constraints matter, and what done looks like. Then I write up a concrete plan.

02

Building

Weeks 2–6

I work iteratively and get something running early. You see progress as it happens, and we refine from there.

03

Delivery

Week 6+

You get working software, documentation, and whatever handoff makes sense. I can stick around for ongoing work or help your team take over.

Have a
project in mind?

Tell me what you're working on. I typically respond within a day.