Boutique AI engineering practice
A boutique practice helping teams architect, build, and ship products that hold up in production — with a hands-on engineer who’s been building AI-native products since the early days.
For 20 years I’ve built and led production software — owning the whole arc from architecture and systems design through shipping, measuring, and operating real products. That work has spanned early-stage startups, big tech, and F500 programs.
Today most of it is AI-native: agents, retrieval, LLM-powered systems, and AI-assisted development as a primary way of shipping. That work rests on a deep search and retrieval background — two decades on the relevance, ranking, and context-assembly problems that turn out to sit at the foundation of how modern AI systems actually work. Laguna Labs is where that range comes to your team — as a fractional technical leader, an AI-native engineering partner, or a strategic advisor.
The shape of the work is generalist: systems thinking, architecture, and product judgment that travel across domains and stacks. The depth comes from that search lineage — the same discipline that, under a new name, now underpins context engineering and agentic retrieval.
How I help
Embedded as the technical anchor on a high-stakes initiative — setting architectural direction, owning delivery end-to-end, making the build-vs-buy and sequencing calls, and raising the bar on review rigor and operational practice. Fractional CTO or principal-engineer leadership for the parts of your roadmap that most need it.
Designing and shipping the AI parts of your product — agents, retrieval/RAG, LLM features — and helping your team adopt AI-assisted development the way effective teams do it: agent harnesses, eval-driven workflows, and patterns that accelerate delivery without trading away review discipline or test quality. Not “vibes coding.” Production engineering, made faster.
Architecture reviews, AI and product roadmaps, build-vs-buy decisions, and senior-engineer mentorship. A considered second opinion from someone who's shipped the kind of thing you're scoping.
The Product Engineer
The industry recently put a name to a role a lot of builders have long played: the Product Engineer. Someone who pairs real engineering depth — systems thinking, architecture, the ability to build hard things — with genuine product instinct, design sensibility, and ownership of the whole lifecycle, from problem definition through what real users actually do with the result.
That’s been the shape of my work for years, well before the label existed:
The practical value is one person who can sit in both the architecture review and the product review and add signal in each — thinking in systems and in outcomes, scoping the right thing and then building it. That combination of systems depth and product judgment in the same person is what a fractional technical leader is for.
Where search meets AI engineering
I started out inside CNET — the team where Apache Solr was born — and spent two decades on search and retrieval: Lucene and Solr at large scale, language models for retrieval (Word2Vec, OpenNLP) years before that was common, and now hybrid dense+sparse retrieval and agentic systems in production.
That lineage turns out to run straight into the center of modern AI. Every serious LLM or agent system depends on what goes into the context window — which documents, which tools, which memory, ranked and assembled under a token budget, in milliseconds. That’s retrieval. Chunking, embeddings, ranking, relevance evaluation, recall-versus-precision tradeoffs — the search field worked these problems for two decades before anyone said “RAG.”
So when a team hits the familiar wall where an agent works in the demo but not in production, the cause is often a retrieval and context problem in disguise. That’s well-worn ground for me, and it tends to be where I can help a team move faster.
You don’t need a search problem to work together — but if AI is anywhere on your roadmap, understanding those foundations is often what separates a demo from a product.
Across stages & environments
Where I’ve gone deep
If your domain is messy, regulated, high-volume, or all three, it’s familiar ground.
Results in production
Latency roughly halved
with new Rust-based systems on one of the company’s highest-revenue product surfaces.
Tens of millions in incremental annualized earnings
attributable to those improvements.
Hybrid dense + sparse retrieval
shipped as the core production path and the reference architecture other teams build on.
An LLM-as-judge evaluation platform
now used as the standard offline-eval tool across search teams.
An end-to-end agentic optimization system
that automates the MLE loop — hypothesis generation through scoring and candidate selection — and has produced live production A/B candidates.
These are measurable outcomes, and I’m glad to walk through how they came about.
How I work
Outcomes over output. Scope to the problem worth solving, not the feature list. Product judgment is part of the engineering.
Evals before opinions. If quality can’t be measured, it’s a guess. The measurement comes first.
Production is the only real demo. Optimize for what survives real traffic, latency budgets, and SLOs — not what looks good in a notebook.
Hands on the keyboard. Leadership by building, shipping alongside the team rather than from the sidelines.
Leverage your people. The best outcome is a team that’s stronger afterward. Mentorship and knowledge transfer are part of the deliverable.
About
Based in the Los Angeles & Orange County area, working remotely with teams anywhere. M.S. and B.S. in Information Management from Syracuse University. AWS Certified Solutions Architect – Professional. Twenty years and counting building products at the intersection of systems, product, and AI.
Let’s talk
Whether you need a fractional technical leader, help shipping the AI part of your product, or a considered second opinion on where you’re headed — get in touch.
hello@lagunalabs.ai