AI Agent · Mobile · 2C · Founding Designer · 2026

Nestify — Vibe Building an MVP
that People Get

Product
Nestify — Family AI Agent
Platform
iOS / Android
My Role
Founding Designer
Timeline
6 Weeks
Team
PM + Multiple Eng

Nestify helps busy families manage household life with an AI agent. When I joined, the team had a working Vibe-Coded MVP, but one that was hard to use and failed to show its value. In one month, I rebuilt the UX, redesigned the entire product, and rewrote how design and engineering work together.

01 / Context
1. A strong idea but held back by the experience

The team had shipped a Vibe-Coded MVP. The core idea was solid: an AI agent helping busy families stay coordinated. But the experience was rough — hard to navigate, visually inconsistent, and missing the moments that would help users feel what made Nestify different.

My job: make it usable, make it feel designed, and make sure users understand in the first two minutes why this isn’t just another calendar app.

02 / Design Process
One month. Three phases.

I worked in three sequential sprints — structure first, system second, surface third.

01
UX Audit & Restructure — Week 1
Mapped all existing flows, identified friction and broken patterns, rebuilt IA and core journeys from scratch.
Deliverable: Revised IA, annotated flow maps, prioritized problem list
02
Design System — Week 2
Built a new design system from scratch: typography, color, spacing, components. Consistent enough to apply across the full product in two weeks.
Deliverable: Full Figma design system — tokens, components, usage guidelines
03
Full Redesign & Rebrand — Weeks 3–4
Applied the new system across every screen. Redesigned every key flow. Rebranded the product — premium, intentional, distinctly AI-native.
Deliverable: Full hi-fi redesign, engineering-ready
New Nestify Design System
Selected redesigned screens
03 / Case Study — Onboarding
The first two minutes had to answer two questions at once

Even after a full redesign, we had a deeper problem. Research revealed users were failing on two dimensions in their first session.

VALUE
“This is just a calendar app”
Users couldn’t distinguish Nestify from existing tools in their first session. The AI agent’s advantages weren’t visible from the start.
USABILITY
Users didn’t know how to use the AI agent
The AI features were present, but users had no mental model for interacting with them. They defaulted to treating Nestify like a manual calendar.
How might we design an onboarding that teaches users how to use the agent AND shows them its value, within the first 2 minutes?

Because we needed real user data to validate the right direction, we shipped an onboarding flow as a test. The onboarding became our fastest path to learning.

03 / Onboarding — Design Process
Research → Ideate → Prototype → Adjust → Document
R
Research
Reviewed session recordings from the existing MVP. Mapped exact moments where users stalled, misunderstood the AI, or dropped off.
Session recordingsFlow mapping
I
Ideation
Explored progressive disclosure vs full setup upfront. Landed on collecting meaningful user data first, then using it to immediately demonstrate what the AI can do — creating a first magic moment.
Whiteboarding
P
Prototype
Built a working prototype in VS Code + Claude — interactive, realistic, testable on real devices. Not a Figma click-through. Real interactions, real transitions.
VS Code + ClaudeStitchCursor
Adjust Hi-Fi
After prototype testing, refined the high-fidelity design. Applied the design system throughout. Documented every state — success, error, loading — for engineering handoff.
Figma
A
Document
Wrote detailed documentation covering the full flow — which steps are final hi-fi, which have outstanding dependencies, and clear notes on what's ready vs. what needs backend support.
NotionFigma annotations
D
Family profile setup
Step 1 — Family setup
Calendar connect screen
Step 2 — Connect calendars
Voice intro / AI demo
Step 3 — Meet the agent
04 / AI-Assisted Workflow
Solving the last-mile design problem

The redesign exposed a structural friction: engineering could get to 60–70% of the intended design quickly, but the final 30–40% — the pixel-level polish that makes a product feel designed — was painfully slow and required constant back-and-forth.

What if the designer owned the last 30–40% of frontend implementation?

As Vibe Coding tools matured — especially for frontend — I proposed a new model: I take ownership of the final implementation layer. Engineers bring it to ~70%, I take it to 100%. I documented the exact boundary, aligned with the founder and engineering team, and we shipped it.

Design
FigmaFigma AI
Full hi-fi with complete design system. Every component, state, and interaction specified.



‍
Prototype
VS Code + ClaudeStitch
Working code prototypes, testable on real devices. Engineering handoff starts from something real.


‍
Eng 0 → 70%
CursorVibe Coding
Engineering implements core functionality and structure. Fast, focused on logic and data — not pixel precision.


‍
Designer 70 → 100%
CursorClaudeCodex
I take the frontend the last mile — spacing, animation, micro-interactions, visual polish. What used to require 20 rounds of back-and-forth, now closed in one pass by the person who can actually see the difference.
Document
Notion
Full documentation — boundaries, responsibilities, tools, prompt templates — so any designer can follow the same process.
Design intern onboarded to full productivity in under 1 week.
1.5 weeks
2 days
Feature delivery time
1 week
1 day
Engineering handoff cycle
weeks ramp-up
<1 wk
Intern onboarded
06 / What's for future
As a growing team, all actions need to be scalable

xxxx

IN PROGRESS
Login — Calendar OAuth

Login-to-calendar redirect pending Apple Calendar and Google Calendar API integration. UX flow designed; implementation needs backend.

IN PROGRESS
Calendar Connection — Onboarding

The “Connect your calendar” step. All connection states (success, error, loading) fully designed. Connection action needs backend support.

IN PROGRESS
Voice Interaction Animation

Waveform animation during voice input fully specified in documentation — visual behaviour, timing, states. Implementation pending.

IN PROGRESS
LLM Output — Live Data

AI agent response after onboarding designed with example content. Rendering real user-specific output requires database connection.

06 / Reflection

Good design isn’t just what you make — it’s how you close the gap between what you intended and what actually gets built.

The best insight from this project wasn’t a UX pattern. It was realizing that designers — not engineers — are often best positioned to own that final implementation layer. Because we’re the ones who can see what’s still wrong at 70%.

Building the workflow to make that possible, and documenting it so the whole team can repeat it, was just as important as the redesign itself.