Marketing Machine ADK
🚀 Inspiration Marketing is broken. Small teams and startups are buried in tools, stuck doing everything manually, or paying agencies thousands just to launch a single campaign. I’ve been in those shoes—wasting time trying to pull together research, creatives, copy, and video with limited resources. That frustration is what pushed me to build Marketing Machine ADK: an AI-driven platform that creates complete marketing campaigns in minutes, not months.
💡 What It Does Marketing Machine ADK is a fully working AI marketing assistant that pulls together all the pieces of a campaign—research, creative ideas, visuals, scripts, and even videos—in 2 to 3 minutes. Users just plug in their business info, and my multi-agent system delivers:
Market Research: Real-time Google Search analysis
Campaign Concepts: 2 unique ideas crafted via Grok API
Visual Design: Instagram-style posts using Imagen 3.0
Video Scripts: Short-form cinematic scripts ready for Veo
Video Generation: 5-second marketing clips produced with Veo 2.0
The whole thing is live and running at: https://adkchl.web.app
🛠How I Built It I designed the platform using Google’s Agent Development Kit (ADK), creating a multi-agent system where each AI agent has a clear role:
Marketing Agent (root): Gemini 2.5 Flash + Google Search for research
Research Specialist: Structures the raw data
Creative Director: Grok API for campaign generation
Visual Concept Agent: Imagen 3.0 for visuals + captions
Script Writer Agent: Generates video-ready copy
Veo Generator Agent: Produces videos via Veo 2.0
Tech stack:
Frontend: Vanilla JS + Firebase Auth + Hosting
Backend: FastAPI on Google Cloud Run
AI Stack: Gemini 2.5/1.5 Flash, Grok, Imagen 3.0, Veo 2.0
CI/CD: Cloud Build for automated deploys
The key was getting ADK to behave—keeping tools isolated and making sure only the root agent accessed things like Google Search, while sub-agents handled their roles cleanly.
⚙️ Challenges Learning ADK: It’s powerful but has a learning curve. Tool isolation, session tracking, agent roles—it all had to be done right or the whole flow broke.
Agent Coordination: Syncing six agents across tasks without stepping on each other was tricky, especially during async operations.
Video Generation: Veo 2.0 isn’t plug-and-play. I had to figure out how to optimize scripts and manage longer video processing reliably.
Frontend Integration: Making the UI responsive while juggling async steps (research → creative → video) and catching errors properly.
Security: Keeping API keys safe, implementing Firebase Auth, and deploying in a way that won’t blow up in production.
✅ Wins I’m Proud Of It Works: It’s live. Real users are generating campaigns with actual videos in under 3 minutes.
Deployed & Stable: Running on Google Cloud with solid uptime.
ADK Expertise: Built a fully compliant multi-agent system using ADK and learned a ton in the process.
Multi-AI Integration: Tied together 5 different AI services into a single pipeline that feels smooth.
Real-World Utility: I’ve generated legit campaigns for businesses like tent rentals and user research tools—it’s not just a demo.
Optimized Performance: Workflows run in parallel, and I’ve tuned API usage to keep everything snappy.
đź“š What I Learned Multi-Agent Architecture: Designing agents that cooperate well takes more than just duct-taping APIs together.
Cloud Ecosystem: I got deep into Cloud Run, Firebase, and CI/CD workflows.
Video AI: Prompting matters a lot. Good video scripts don’t just happen—they’re crafted.
Going Beyond MVPs: There’s a big difference between a cool AI demo and a system that handles real users, edge cases, and uptime.
Quotas & Limits: Learned how to handle rate limits gracefully and keep the user experience smooth under pressure.
🔜 What’s Next Longer & Smarter Videos: 15–30 seconds, more variations, user branding
Global Campaigns: Multilingual generation + region-specific strategies
Analytics: Built-in A/B testing, ROI tracking, optimization tips
Advanced Targeting: Demographic and industry-based personalization
Automation: Scheduled campaign runs, batch generation, CRM integrations
Team & Enterprise Tools: Brand guideline support, collaboration, fine-tuned models
Marketing Machine ADK is just getting started. I’m building toward a future where AI takes care of the busywork, so teams can focus on strategy and growth—not wrangling tools or designing Instagram posts.
Log in or sign up for Devpost to join the conversation.