I turn scattered context into working software: agentic workflows, internal tools, automation, and product systems people can actually use.
Most AI work does not fail because the model is weak. It fails because nobody captured the context, workflow rules, handoffs, review loops, and ownership model. That is the layer I like building.
Agents with durable context, task state, tool access, and judgment boundaries.
Dashboards, admin panels, intake flows, and control surfaces for repeated work.
Avatar, voice, video, content, and review pipelines using modern AI media APIs.
React, TypeScript, SQLite, Supabase, PocketBase, APIs, and pragmatic glue code.
Public and safely describable examples that show how I think, build, and turn ambiguity into working systems.
A live personal execution system with onboarding, payments, email, app flows, and AI-assisted planning concepts. Built as a tool for turning messy input into action.
A private pattern for agent memory, project state, decisions, incidents, follow-ups, and dashboards so AI work has continuity.
Contracted AI systems work in a real operating environment. Public details stay high-level because of NDA boundaries.
Control panels and experiments around AI avatar video, voice, generated media, content review, and API-driven production loops.
A playable multiplayer browser game showing 3D, TypeScript, physics, and real-time interaction.
Play live open_in_new
A focused tool for turning confusing pay questions into a simple review flow.
A compliance-oriented prototype for making structured review easier to follow.
Good AI work is not just prompts. It is context capture, interfaces, review loops, permissions, data shape, failure handling, and the human confidence to use the thing when the work gets noisy.
Hospitality and customer-facing work taught me how to translate messy human context into tools people can trust.
Most valuable systems start as fog. I can explore, prototype, test, and tighten until the shape appears.
I do my best work with room to own the problem, make decisions, and show progress through working artifacts.
AI does not fix chaos by magic. It needs rules, memory, context, and maintenance to become useful infrastructure.
My best work is turning unclear operational problems into concrete systems: interfaces, workflows, data shape, automations, and agent loops that people can trust. Available for aligned full-time remote roles, contract work, and consulting where practical AI tooling matters.
Send me the context. I will tell you what I would build first, what I would leave alone, and where a useful version probably starts.