Twenty-five years
of building
technology that ships.
I'm an AI product and engineering leader. This is where I write, run experiments, and think out loud about the work — in public, in measured terms, and on my own terms.
AIOps Intelligence
A unified command center for AI agent operations — real-time token spend, security alerts, P95 latency, and cross-platform activity across Agentforce, GitHub Copilot, Vertex AI, and more.
A practitioner who still writes the code.
Twenty-five years across architecture, product leadership, and applied AI — for global enterprises and for two-person teams. The instinct is the same in both rooms.
I lead teams that ship AI products. My early career was in distributed systems, identity, and large-scale data — the unglamorous plumbing that makes everything else possible. The last decade has been about putting machine learning, then generative AI, into production where it has to earn its keep alongside the rest of the stack.
Today I split my time between hands-on engineeringon the experiments you'll find in Labs, advisory workwith founders and platform teams, and writing about what I'm learning. I don't sell consulting hours as a product — but I do take on a small number of collaborations each year where the problem is interesting and the team is real.
The site name — NuvTek — is a contraction of nuevo tech: new technology, built by someone who remembers the old technology. The aesthetic of this site is the aesthetic of the work: measured, declarative, and as plain as it can stand to be.
Three threads, one craft.
Not a services menu — a description of where the time actually goes. The threads inform each other; none of them stand alone.
AI product leadership
Scoping, sequencing, and shipping LLM and ML capabilities inside larger products — with the evaluation discipline to know whether they actually moved a number.
Enterprise architecture
Identity, data, and platform decisions for systems that have to run for a decade. The right call is usually the boring one stated precisely.
Building teams that ship
Hiring, structure, and the rituals that turn three good engineers into a working group. Less ceremony, more written decisions.
Things I'm actually building.
A small, honest list. Live experiments, betas open to a few people, and one or two that are still in the notebook. Status tags reflect today's truth, not the marketing pitch.
AIOps Intelligence — a command center for AI agent operations.
Real-time visibility across Agentforce, GitHub Copilot, Vertex AI, AWS, and ServiceNow — token spend, security alerts, P95 latency, and cross-platform activity in a single view.
Ask the site — an answer agent over my writing.
A small retrieval agent that answers questions about my background, articles, and projects. Built on Claude with a hand-tuned context set. Try it via the badge in the corner.
Eval Studio — a structured eval harness for LLM features.
A workbench for defining evaluation sets, running prompts at scale, and grading rubrics without writing glue code. Currently in a closed beta with two design partners.
Latency Mirror — a visual profiler for agent traces.
A reading tool for multi-step agent runs. Renders the trace as a waterfall, surfaces the costly hops, and lets you replay a step with a different model. Coming this quarter.
Coming soonNotes from the work.
White papers, field notes, and lessons learned. Observations on AI, architecture, and what it takes to ship technology at scale.
From AI Curious to AI First: A Practical Roadmap for Engineering Managers
A field-tested maturity model for engineering managers adopting AI — four stages from Reactive Support to Governed AI-First, with a use-case selection checklist, metrics framework, and hard-won lessons.
Read article →
PR Impact Analysis: What It Is and Why It Changed Our Sprint Cadence
PR impact analysis moves risk assessment upstream into pull requests, automating the coordination overhead that justified sprint batching. This decouples release cadence from calendar intervals, enabling teams to ship when code is ready rather than when sprints end.
Read article →Evaluation is the product. Everything else is decoration.
Why every AI feature I ship now starts with the eval set, not the prompt. A short brief on what changes when you reverse the order.
Read article →A reference architecture for LLM features inside enterprise SaaS.
Identity, audit, fallback, evaluation, and cost — the five concerns that an LLM feature in a regulated product has to answer on day one.
Read article →I retired a six-year-old service this quarter. Here is what it taught me.
Notes from decommissioning a system I designed in 2020. The parts that aged well, the parts that did not, and the line between the two.
Read article →Specs are written for the team that has to maintain the work.
A short argument for spending more time on the document and less time on the meeting that produced it. With a template.
Read article →If the work resonates,
let's talk.
I read every message. Advisory engagements and interesting collaborations get a same-week reply; everything else gets a reply when the calendar allows.
“Driving technical excellence and strategic growth.”
Let's collaborate to build high-impact solutions for what comes next.