Ezequiel Rodriguez — Agentic Product Designer
Visualine: AI-powered audience intelligence for creators and social teams
Role: Product Designer, end to end · Timeline: 2025 · MVP · Team: Founder + engineering team
Visualine is a B2B SaaS platform that pulls comments from YouTube, Instagram, Threads, and TikTok into one dashboard. AI tags sentiment and themes, surfaces trends before they peak, drafts on-brand replies, and flags what needs moderating, so insight, engagement, and moderation happen in the same place instead of three.
The problem
Creators and social teams get thousands of comments across YouTube, Instagram, Threads, and TikTok. Their tools bury the useful feedback under noise, report vanity metrics instead, and generate replies that read like a template. Teams spend hours retyping the same answer while a trend forms and passes.
Discovery & UX strategy
Defined the target users, creators and social teams, and their primary goals: save time, increase engagement, stay recognisably themselves. Mapped the MVP journey (comment ingestion → insights → AI replies → moderation) and the key user journeys before committing to any screen.
Information architecture
Data volume was the main risk, so I set the page hierarchy first (Dashboard, Unified Inbox, Analytics & Insights, AI Reply Composer, Settings) with a sitemap and desktop-first layout priorities that keep the highest-signal information one click from the dashboard.
UX wireframes
Designed low-fidelity wireframes for 10+ core screens: dashboard, unified inbox, AI reply composer, insights feed, moderation, and settings, each with its click paths, transitions, and empty states. Logic and hierarchy were settled before any visual design started.
UI design & design system
Built the visual layer on an adapted Untitled UI foundation, extending it with tables, filters, dashboards, insight cards, and AI interaction states, including empty, loading, and error. After that, adding a data-heavy screen meant assembling existing parts rather than drawing a new one.
Key product flows
Designed five flows with full states, feedback, and failure handling: the Unified Inbox, AI Trend Detection, the AI Reply Assistant with human-in-the-loop approval, Moderation & Hygiene, and a Weekly Insights Digest.
Prototype & handoff
Built clickable prototypes covering the full MVP flow, then prepared annotated Figma screens, component usage notes, responsive guidelines, and interaction explanations, so developers could build from the file instead of asking me what a state was supposed to do.
Challenges & solutions
- Designing for AI without losing trust — Challenge: Avoiding black-box AI behavior and generic replies. Solution: Human-in-the-loop workflows where users review, edit, and approve all AI output.
- Managing data density — Challenge: Large volumes of comments and insights risked overwhelming users. Solution: Strong hierarchy, filters, saved views, and progressive disclosure.
- MVP constraints — Challenge: API limits, latency, and AI cost constraints. Solution: Fallback behaviors, async loading states, and copy-to-clipboard flows where auto-posting was not possible.
The outcome
Delivered the MVP end to end: clickable prototype, design system, and handoff documentation. The founder is taking it into pre-seed fundraising and beta onboarding as a growth-intelligence platform.
- 10+ core screens wireframed and shipped to hi-fi
- 4 platforms unified into one inbox
- 5 product flows designed with full states
Key takeaways
- People trust AI output they can edit before it goes out
- Ranking comments by signal did more for legibility than adding filters
- Settling the page hierarchy first made every later screen cheaper to draw
- Copy to clipboard beat auto-posting wherever the API would not cooperate
All work · Contact ezecrodriguez22@gmail.com