AI-native product designer and workflow transformation leader.
I turn complex workflows into trusted product systems, then build the AI-native tools, routines, and playbooks that help teams work differently for good.
Thirteen years of transformation, now AI-native.
Before AI
Transformed messy, high-stakes workflows into trusted product systems across enterprise B2B for over a decade.
At Optera
Led product design across five climate-tech platforms, turning spreadsheet and services-heavy work into reusable workflows, research rituals, feedback systems, and product infrastructure.
AI-Native Now
Operating through ChatGPT, Claude, Claude Code, Codex, GitHub workflows, versioned skills, scheduled routines, Lovable, TanStack, Notion, Obsidian, Figma Make, and Figma MCP.
What changed
Discovery, prototyping, validation, documentation, and enablement now happen inside one connected operating system.
Selected artifacts.
Optera
Climate-tech2021 — 2026
Five years turning high-stakes climate workflows into scalable product systems: automated data intake, Feedback OS, supplier response management, design systems, and research rituals.
FieldRules
AI Governance2026 — Present
AI governance product and operating archive. BECAUSE elicitation, Reasoning Health Score, eval-driven development, provenance-first schema, design system, and a 2,265-page Notion/Obsidian artifact system.
Helm
Strategy / OKRs2026 — Present
AI-native OKR and strategy platform work. Product surfaces, design system, Claude skill suite, current-state audits, AI review gates, and pull-AI vs push-AI interaction principles.
GridPath
Public-data prototype2026
Public-source grid intelligence prototype connecting EIA, LBNL, NOAA/NWS, EPA, Census, WattTime, HIFLD, PowerOutage.us, and Open-Meteo signals.
Quiet Hours
Energy timing2026
Location-aware energy story using GridPath, EIA, CAISO, NOAA/NWS, EPA eGRID, OpenEI rates, NREL ResStock, Census ACS, and permissioned location.
Between Nowhere
Learning system2026
FCC Amateur Radio Technician question pool turned into a mobile-first learning product.
Helm as proof: design the operating system, not just the interface.
Helm shows the shift from designing screens to designing the operating model around product work: skills, routines, audits, AI surfaces, review gates, and documentation moving together.
See Helm in the work section →Customer-loop synthesis, surface audits, design-system updates, prototype-to-ticket workflows, and reusable Claude Code skill patterns.
Recurring current-state checks, drift audits, quality reviews, and source-of-truth documentation.
Proactive coaching, on-demand status checks, Slack-first handoffs, AI-prepared reviews, and human judgment gates.
Not a tool list. An integrated workflow.
Discovery, prototyping, validation, documentation, and enablement live inside one connected stack. Every artifact carries provenance. Every routine compounds the next.
- ChatGPT
- Claude Chat + Claude Cowork
- Claude Code research skills
- Lovable
- Codex
- Claude Code
- GitHub PR/review skills
- TanStack Start
- Notion
- Obsidian
- Claude memory skill
- GitHub-sourced Claude skills
- Recurring coherence routines
- Lovable
- Figma Make
- Figma MCP
- Claude Code frontend skill
- Lovable + Codex + TanStack loops
How I transform workflows.
A six-step system honed across enterprise B2B, climate-tech, and AI-native product work.
- 01
Diagnose the messy workflow
Map the friction, the silent handoffs, and the human logic hiding inside the work.
- 02
Find the highest-leverage bottleneck
Identify the single move that unlocks downstream speed, trust, or quality.
- 03
Build a working tool or prototype
Live data, real interactions, real edges — not a clickable mockup.
- 04
Validate with users, data, and quality gates
Synthetic and adversarial evals, public-source data checks, human review, measured trust improvements.
- 05
Document the playbook
Provenance, decisions, and reasoning captured so the pattern travels.
- 06
Teach the team and scale the pattern
Routines, skills, and shared standards that compound across the org.
Built teams. Shipped systems. Taught patterns.
VP / Director-level design leadership
Built and mentored design teams across stages
Built design systems and shared quality standards
Led workshops and cross-functional alignment
Translated technical complexity for non-technical teams
Shipped practical, principled AI adoption guardrails
The pattern people name after working with me.
“product designer and a product thinker in the same person”
Michael Koenig
Managed Melanie directly
“one of the most creative and collaborative leaders”
Jenny Jones
Cross-functional partner
“benchmark for excellence that I still reference”
Hilary Rallo
UX engineering partner
“champion for introducing AI into our workflow, across multiple teams”
Gautami Chennur
Direct report
“one of the most exceptional design leaders I’ve ever collaborated with”
Ty Colman
Executive partner
“fierce advocate for her team”
Katie Oakes
Leadership teammate
How I think about the next era of design.
Headless does not mean designless
When a platform is mostly consumed by agents, design does not go away. It moves into contracts, defaults, permissions, feedback loops, tool descriptions, source health, and the parts of the product humans may never directly touch.
Read essayProduct developmentThe SDLC is changing under our feet
When designers can shape schemas, prototype against real data, and ship production-level features, the path to MVP stops being a long relay race and starts becoming a tighter loop of judgment, evidence, and build.
Read essayPoint of viewDesign is moving closer to the work
AI-native design is not about making designers faster at making screens. It is about collapsing the distance between judgment, evidence, implementation, and the product itself.
Read essayOperating modelTeams need one living body of knowledge
The AI-native team does not win because it has more tools. It wins because its tools can reason from the same evolving truth.
Read essayAI-native practiceAutomating ideation with science, not slop
The useful version of AI-assisted product improvement is not a machine that generates more ideas. It is a system that can pressure-test ideas against UX heuristics, behavioral science, HCI research, product evidence, and the realities of machine learning.
Read essay