$5B Enterprise Ad Sales Platform
[1]
Consolidating a 5-System Legacy Stack into a Unified Operating Engine
If you'd like to learn more, please get in touch.
Role
Solo Design Engineer
Team
Ad Sales Technology
Tools
Figma, Claude Code, Vercel, GitHub, AI-Assisted Pipelines
Context
[2]
The problem
Paramount’s ad sales ran across 5 disconnected legacy systems. Sales reps manually rebuilt advertiser records across tools, causing data fragmentation and corrupting revenue forecasting across a $5B business.
The solution
As a solo design engineer, I collapsed five systems into one platform; six flows I designed and built myself, on an AI-assisted pipeline that cut design-to-prototype from weeks to days and made the advertiser record a single source of truth
Accelerated Technical Alignment
[3]
Shifted from static handoffs to production-like prototypes, collaborating directly with engineering leads in live sessions to validate technical feasibility and edge-case data flows in real time.
Trade Offs & Architectural Decisions
[4]
Every design decision in this platform was driven by three simple questions every sales rep needs answered:
Solving Data Quality at Entry
The Decision: Block duplicate records before they are created.
How It Works: Built real-time record checking into the initial input fields across 21 agencies and 7 holding companies.
Why It Matters: Preventing bad data at entry protected downstream revenue reporting across high stake ad deals.
Progressive Disclosure for Deep Taxonomies
The Problem: Media buying involves dozens of parameters (network restrictions, time blocks, exclusions) that usually result in massive, unreadable forms.
The Trade-Off: Instead of showing 60+ options at once, I used tag-based progressive disclosure (CBS News ✕, 5AM–9AM ✕).
Key UX Choices:
3-Step Task Chunking for setup
Inline Tags for instant visual auditing
Collapsed Tail Lists/ default views to keep the UI clean.
Predictive Feedback & Tiered Error Handling
The Problem: Media buying involves dozens of parameters (network restrictions, time blocks, exclusions) that usually result in massive, unreadable forms.
The Trade-Off: Instead of showing 60+ options at once, I used tag-based progressive disclosure (CBS News ✕, 5AM–9AM ✕).
Visual Health States: Uses segmented progress bars and semantic colors (Green/Yellow/Red) to show deliverability status instantly.
Prescriptive Fixes: Surfaces direct actions ("Choose Peak Placement", "Expand audience") when deliverability drops.
Tiered Warnings: Uses soft inline banners for minor risks and blocking modals only for critical budget/targeting conflicts.
Narrow targeting rules can accidentally starve campaign delivery, causing missed revenue targets
→
Engineered live data feedback components to catch issues early.
Inline Checkpoints over Review Pages
Traditional review pages are passive text dumps users blindly click through. I replaced them with a progressive checkpoint architecture
Ending Thoughts
[5]
In next steps I would instrument flow tracking and run A/B testing to capture task completion speeds, drop-off rates, and clear ROI metrics for executive leadership.
This project helped me learn to use live code as the ultimate alignment tool.
High-velocity prototyping bought instant credibility with operational teams. Shipping live URL prototypes eliminated endless debate over static mockups and exposed real data constraints in days rather than sprints.
Footer
[6]