AI-Native Learning Paths

Developed the curriculum vision and strategy for Skillsoft's AI-native learning paths — defining how developers learn to build with AI tools from the ground up, not as an add-on, but as the primary mode of development.

The core thesis: learners who start with AI-assisted workflows develop an AI-integrated mindset from the beginning, rather than having to unlearn traditional approaches after mastering basic programming. AI handles syntax and boilerplate while learners focus on architectural thinking, problem decomposition, and system design — compressing learning timelines while improving conceptual depth. The goal is workforce-ready graduates who are already fluent in professional AI-assisted development workflows.

Codecademy AI Builder

Contributed to the product strategy and instructional design philosophy for Codecademy's AI Builder — a feature that inverts the traditional learn-then-build sequence. Learners start by building something they want to make; the AI generates both the project and a personalized learning path explaining how it was built.

The key design challenge was ensuring the experience built durable mental models rather than surface-level familiarity with AI-generated code. The experience is grounded in the Socratic method: the AI asks questions and gives strategic nudges rather than handing over answers — because research on AI in education shows that simply providing answers makes it harder for learners to retain information on their own.

AI Test Agent

Leading the curriculum vision for AI Test Agent — a new assessment type that uses an LLM to evaluate learner code and text responses at runtime. Traditional checkpoint tests are powerful for syntax validation and straightforward code snippets, but they have real limits: they can't assess large-scale code structure, evaluate non-code responses, or handle problems where multiple valid solutions exist.

AI Test Agent fills those gaps — enabling assessment of complete application architecture, written explanations of design decisions, and flexible code evaluation where the goal is achieving an outcome rather than matching a specific implementation. The work required thinking carefully about where AI evaluation adds genuine value versus where traditional tests remain the right tool.

Agentic Coding in the Learning Environment

Currently collaborating with engineering to bring agentic coding tools into Codecademy's interactive learning environment — enabling learners to work with AI coding agents directly inside the platform, without switching context. The goal is to make working with AI agents a first-class part of the learning experience, not something that happens outside the curriculum.

This work sits at the intersection of platform engineering and curriculum design: figuring out not just how the tool works technically, but what learners need to understand to use it effectively and develop judgment about its outputs.

AI-Assisted Content Production System

Designed and led a two-phase system for integrating AI into the curriculum content production workflow. The insight driving it: AI agents are only as good as the context you give them. Before building AI tooling, the foundation has to be right.

Phase 1 was a documentation sprint — updating and standardizing curriculum documentation so it could serve as reliable, structured context for AI agents. Phase 2 built the AI-assisted content creation tools on top of that foundation, with team-wide onboarding built directly into the workflow itself. The result was a 50% reduction in production time across the team, without sacrificing content quality.

This project also informed my broader thinking about learning-driven development: the onboarding for the AI tools was built into Claude Cowork using the onboarding skill, so instructional designers learned how to use the tools while they were using them to do their jobs.

Teaching