AI product and full-stack engineer

I build reliable AI products and the systems that make agents useful in production.

I'm Rebecca, an Oregon-based engineer with 14 years of experience turning ambiguous ideas into dependable products. My work spans agent architecture, context and retrieval, full-stack product experiences, evaluation, observability, and developer platforms.

  • 6xgrowth in daily use of a production agentic product
  • 35%increase in in-app customer satisfaction scores
Profile photo

How I got here

From product platforms to production AI

I've spent 14 years building software products from the interface down through the systems behind them. Earlier in my career, that meant frontend architecture, design systems, developer tooling, and shared platforms used across large engineering organizations.

As AI became part of the products I was building, my focus shifted to a harder question: how do you make systems built on probabilistic models behave like dependable software?

Today I work at that boundary, designing agent architecture, context management, retrieval, evaluations, state machines, observability, and the product experiences around them.

Selected Work

Sudowrite Agentic Chat

Sudowrite Agentic Chat

An AI writing assistant that connects conversation with manuscript tools, preserving the author's intent through revisions, approvals, and interrupted work.

Plant Pairings

Plant Pairings

A companion planting field guide and interactive 3D garden planner with cited pairing guidance, seasonal filters, and USDA zone information.

Bend Neighborhood Development Agent

Bend Neighborhood Development Agent

A map-centered research agent that connects public permit, planning, property, and geographic data to make neighborhood development progress easier to understand.

Areas of Expertise

Product, platform, and agent architecture for reliable systems teams can extend with confidence.

Agent Systems

Production agents built for reliable, measurable behavior.

  • ArchitectureOrchestration, multi-agent systems, state machines
  • Tools & ContextTool calling, MCP, retrieval, model routing
  • Quality & SafetyEvaluation, observability, behavioral verification

Product Engineering

Customer needs translated into polished products from discovery through iteration.

  • ApplicationsTypeScript, React, Next.js, React Native
  • Interaction DesignConversational UI, streaming, optimistic experiences
  • Product QualityAccessibility, performance, testing

Platforms & Infrastructure

Shared foundations for reliable delivery and independent teams.

  • Services & DataNode.js, APIs, PostgreSQL, Redis
  • InfrastructureAWS, CI/CD, observability
  • Developer ExperienceShared libraries, tooling, automation

Technical Writing

When to Split Your Turborepo Config into Package-Level Configs

When to Split Your Turborepo Config into Package-Level Configs

Refactor a crowded turbo.json into shared defaults and package-level configs, understand inheritance, and verify dependency ordering and cache behavior.

Cross-Origin Storage Could Make Local AI on the Web Practical

Cross-Origin Storage Could Make Local AI on the Web Practical

Explore how the experimental Cross-Origin Storage proposal could let web applications reuse identical AI model files, with TypeScript examples and a plan for measuring cross-origin reuse.

Can My Browser Connect to This Bluetooth Device?

Can My Browser Connect to This Bluetooth Device?

Try a Web Bluetooth service analyzer to see what your BLE device advertises, which services your browser can access, and where a connection gets stuck.

How to Save and Resume a Streaming LLM Response After a Client Disconnect

How to Save and Resume a Streaming LLM Response After a Client Disconnect

Interrupt a streaming chat response, inspect the chunks saved in SQLite, and reconnect without losing a word or starting a second generation.

Explore all writing

From Bluesky

Here are the things I've been excited about lately.

Work With Me

Looking for an experienced engineer to lead an ambitious product, strengthen an agent system, or create leverage across teams? I'd love to hear what you're building.