Portfolio

I turn messy product problems into clear structure, usable flows, and credible interfaces

About

Dima Ginzburg

Product Designer focused on AI-native and early-stage products

I design agentic experiences - shaping how systems reason, act, and interact with users across flows, not just screens. Moving beyond interfaces, I focus on behavior, orchestration, and generated outcomes - translating product ideas and AI capabilities into clear, usable systems.

Agentic flows
AI systems
Interaction models
End-to-end design

Cover

Skills and tools

/ skills

Product & User Experience

Product strategy
UX system
Product framing
Product narratives
Interaction design

Research & Validation

JTBD
User research
Persona hypothesis
CJM
A/B testing

Delivery & Execution

Prototyping
Wireframing
Micro-interaction
Design systems
shadcn/UI Kit

AI Product Design

AI-assisted UX
Prompt prototyping
Design automation
Agent workflows

/ tools

Design

Figma
Sketch
Adobe XD
Framer
Stitch

AI IDE

Codex
Claude Code
Cursor
Windsurf
1Code

Past projects

/ Past projects

RZD sanatorium and resort booking service cover

RZD • Sanatorium & Resort Booking Service

Turning a complex internal booking process into a predictable and repeatable system

Russian Railways • Enterprise Communicator UX

Designing a component-driven interaction system under legacy constraints

Megamod • Structuring a Multi-Product Narrative

Scaling product complexity through a unified communication layer

Travel Booking Platform • Decision-Driven UX

Helping users move from vague intent to confident booking

/ BUILDS

Deus in Machina

Design System Runtime (DSR) — enforcement + structure + runtime for design systems — focused on token normalization, pattern inference, and UI validation with a fix loop

DEUS-IN-MACHINA

Assistant Workflow

Deterministic AI workflow with Phase → Block → Task → Session, approval gates, and file-based context

GITHUB

Design Token Auditor

AI-powered design token analysis for JSON files with scores, stats, and actionable recommendations

GITHUB

Shimlayer

HITL API layer for last-mile failures in AI-agentic workflows

GITHUB