From Manual Research to AI-Powered Audience Intelligence for a Global Ad Leader

Media planners at a Fortune Global 500 advertising company were spending days manually researching audiences for each campaign. Audience data was siloed across disconnected systems, taxonomy management was a manual spreadsheet exercise, and there was no way to connect brand intelligence, consumer interests, and targetable segments into a single queryable source of truth.

The result: slow campaign setup, inconsistent targeting across markets, and no scalable path to roll out audience capabilities to agencies worldwide.

The client needed a platform that could take a natural language campaign brief and automatically surface the right audiences — replacing manual research with AI-driven discovery, and scaling across 9+ country launches planned for the year ahead.

As Tech Lead, I led cross-functional engineering teams through the design and delivery of an interconnected platform on Google Cloud — from knowledge graph architecture to AI-powered audience matching to enterprise auth.

1. Knowledge Graph — the foundation

   • Built a marketing knowledge graph (Neo4j Aura Enterprise) connecting 70,000+ entities — brands, products, consumer interests, keywords, and 28,000+ targetable audience segments — across 440,000+ relationships
   • Unified data from market research providers, open knowledge bases, and proprietary surveys into a single queryable graph
   • Exposed via a FastAPI Graph API optimized for real-time audience discovery

2. AI-Powered Audience Engine — brief to audience in seconds

   • Built a service that accepts campaign briefs (PDFs or chat Q&A), extracts marketing entities using Vertex AI (Gemini), and matches them to relevant audiences via vector similarity search (PostgreSQL + pgvector)
   • Replaced a process that previously required hours of manual research per campaign with near-instant, AI-driven audience recommendations

3. Segment & Taxonomy Management — replacing spreadsheets

   • Delivered a Segment Manager for building virtual audience clusters by combining filters (age, gender, location, interests, brands) with semantic search powered by Sentence Transformers
   • Built a Taxonomy Manager with interactive graph visualization, replacing manual IAB taxonomy maintenance with a pluggable system supporting CSV imports, custom taxonomies, and graph queries

4. Audience Builder — visual campaign targeting

   • Full-stack application (React + FastAPI) for constructing and visualizing audiences with geospatial heatmaps, overlap analysis, and data exports
   • Integrated into the client's flagship platform via micro-frontend architecture, accessible to planners across the organization

5. Enterprise Security & Multi-Tenancy

   • Designed fine-grained access control (SpiceDB + Cloud Spanner) supporting per-tenant, per-audience-type permissions across global agencies
   • All services on GCP Cloud Run with OpenID Connect auth, GitLab CI/CD, and end-to-end integration testing via Playwright

6. Team & Stakeholder Leadership

   • Led the backend/AI engineering team while coordinating with a separate .NET/Angular frontend team
   • Managed engineers, data scientists, and contractors across time zones — running daily standups, sprint planning, architecture workshops, and client requirements sessions
   • Owned technical decisions on graph architecture, AI integration patterns, and cross-team API contracts

• Manual audience research eliminated: campaign brief to targeted audiences went from days of manual work to seconds via AI-powered matching
• 9+ market rollout enabled: multi-tenant architecture and taxonomy flexibility supported concurrent launches across countries and currencies
• 28,000+ audience segments connected through a 70,000+ node knowledge graph — giving planners a single source of truth for audience intelligence
• Taxonomy management time cut dramatically: interactive graph-based tooling replaced manual spreadsheet-driven processes
• 8 microservices shipped: audience builder, predictive engine, segment manager, taxonomy manager, graph API, geo-translation, integration tests, and auth — all production-grade
• Adopted into the client's flagship operating system, now used by agencies organization-wide

The platform turned audience targeting from a bottleneck into a competitive advantage — enabling the client's agencies to launch campaigns faster, target more precisely, and scale audience intelligence across global markets without adding headcount.