AI Engineering · Agent Systems · Product Architecture

AI systems that can act — and stay under control.

I design and ship agentic products, model integrations, and the engineering platforms around them. The goal is not an impressive demo; it is a useful system with clear authority, durable execution, measurable quality, and a path to production.

Discuss a project View selected work GitHub

10+

Years shipping software

From early client systems to enterprise products and AI tooling.

50+

Applications shaped

Architecture and standards across a large interconnected product suite.

70+

Design-system components

A typed, themed UI platform with documentation and automated releases.

3–4×

Faster build feedback

Achieved through a controlled webpack-to-Vite migration.

Engineering the system around the model

A model call is one component. Reliable AI products also need contracts, context, tools, permissions, evaluation, recovery, and a product experience people can trust.

Agent systems

Autonomous and human-supervised workflows that explore, plan, act, verify, pause, and resume without hiding authority in prompt text.

→ Multi-model routing and bounded agent loops
→ Durable wait/resume and idempotent actions
→ Tool policy, sandboxing, audit, and approvals
→ Evaluation, repair loops, and cost evidence

AI product engineering

End-to-end AI features that connect model behavior to real user workflows, measurable outcomes, and maintainable application architecture.

→ Structured outputs and model integrations
→ Retrieval, context projection, and caching
→ Guardrails and human-in-the-loop UX
→ Observability and quality evaluation

Product platforms

The web architecture and developer experience needed to turn an AI capability into a fast, coherent, production-grade product.

→ Complex Vue, Nuxt, React, and TypeScript UIs
→ Design systems and cross-team standards
→ Incremental legacy modernization
→ API contracts, performance, and delivery tooling

Currently building

Cadrune

A security-first, domain-agnostic framework for embedding governed agent systems into real products and operational workflows.

Active development · public alpha preparation

Provider-neutral coding and task agents composed outside the core
Deny-by-default capabilities, scoped secrets, sandboxing, and audit
Durable execution, clarification, verification, repair, and approval gates
Extension SDKs and language-boundary parity across TypeScript, Python, and Go

Rust runtimeTypeScript SDKPython SDKGo SDKJSON SchemaPlaywright

From product architecture to agent engineering

I have spent more than a decade turning ambiguous product goals into working software. That background matters in AI engineering: models are probabilistic, but the surrounding product, permissions, data contracts, failure handling, and user experience cannot be vague. I work across the boundary between product engineering and platform design — close enough to the interface to understand the user, and deep enough in the runtime to make autonomous execution observable and governable.

Work Experience

Sep 2018 - Present

Lead Frontend Engineer

Aug 2017 - Sep 2018

Jul 2015 - Aug 2017

Aug 2012 - Jul 2015

Full-stack Developer, Founder

ILSER

Notes on AI systems and product engineering

Practical thinking about agents, architecture, delivery, and the engineering discipline around models.

Mar 10, 2025

AI Won't Replace Developers-But It Will Expose the Gaps

Why AI coding assistants make good engineers faster and bad practices more dangerous.

Mar 2, 2025

Design Systems After Validation: Speed Without Chaos

Why early-stage startups should skip design systems-and why scaling products can't survive without them.

Feb 23, 2025

Define the Business Process Before Writing Code

Why startups that skip process documentation waste months on features no one uses-and how to prevent it.

Working together

What I build, how I approach AI work, and where I can be useful.

AI Engineering

What kind of AI systems do you build?

Agentic workflows, multi-model product features, model evaluation and routing, structured-output pipelines, and the platform controls around them. I focus on systems that must do useful work repeatedly — not one-off prompt demos.

How do you keep autonomous agents safe?

Authority stays outside the model. Tools and external writes are deny-by-default, secrets are scoped to operations, risky effects require approval, and every run produces inspectable evidence.

Product Delivery

Can you integrate AI into an existing product?

Yes. I start with the user workflow and business outcome, then map data, failure modes, permissions, and evaluation before choosing the model integration.

Can you modernize a legacy codebase while shipping?

Yes. I have led incremental migrations from jQuery-era applications to modern TypeScript frameworks without freezing feature delivery.

Collaboration

Where can you have the most impact?

Teams moving from AI experiments to dependable product capability; companies building agent infrastructure or developer tools; and products that need both strong frontend architecture and deeper runtime controls.

Are you available for new work?

I am open to selected full-time, contract, and consulting conversations in AI engineering, agent systems, product architecture, and technical strategy. Email me with the problem, current stage, and what success should look like.