AI Product Designer + Product Lead building agentic-native AI workflows.
I design AI products as systems, My work connects human intent, structured context, agent behavior, evaluation, and permissioned learning into workflows people can understand, steer, and trust.
Agentic product workflows
Systems where AI can research, generate, evaluate, and act across multiple steps while keeping humans in control of important decisions.
Context + control layers
UX and harness logic that decide what the AI should know, which rules take priority, what can be remembered, and how corrections change future behavior.
AI-native creative tools
Structured creation environments for image, video, content, and visual exploration that use direct manipulation, presets, references, and review states instead of defaulting to chat.
Research-to-build systems
Artifacts that connect user research, model constraints, workflow architecture, prototypes, implementation, and evaluation into one product loop.
Some production and client repositories are private. I use GitHub to show the public side of the same practice: reusable AI skills, interaction systems, creative tooling, implementation craft, and validation.vc
- Understand model capability — identify strengths, limitations, context needs, and failure patterns.
- Define Agent behavior — goals, states, tools, routing, priority, memory, and escalation.
- Design control — decide how users understand, review, correct, and override AI behavior.
- Build a working prototype or production-facing interface.
- Evaluate with realistic tasks, edge cases, evidence, and implementation constraints.
- Adapt&Multi-Agents System only from feedback the system is actually allowed to learn from.
- Agentic Design for both users and AI agents
- Context engineering and harness behavior
- Human-in-the-loop AI workflows
- Memory and permission models
- Agent evaluation and failure handling
- AI-native creative and operational tools
- Design-to-code systems
Product strategy Agentic UX Workflow architecture Context engineering Evaluation React TypeScript Astro Tailwind Supabase LangGraph Gemini Codex Three.js