Each card shows the real pipeline. Hover a step to see what it does; Play walks a run through, retries included. Open a case study for problem, decisions, validation and result.
Multi-agent lead research and outreach
Text version: Research → Website intel → ICP score → Opportunity → Outreach → QA gate
A six-agent LangGraph system with one typed shared state: Researcher, Website Intelligence, ICP Scorer, Opportunity Finder, Outreach Writer and QA Validator, streamed into a Next.js front end.
On-brand ad creatives at volume, QA-checked by a vision model
Text version: Brief → Prompt → Generate → Vision QA → Regenerate → Approve
A scheduled n8n pipeline: brief → cached prompt → image generation → vision-model QA against written brand rules → automatic regeneration on failure, capped.
Messy documents in, validated records out
Text version: Upload → Extract → Resolve → Validate → Merge
Multi-format intake (vision model for scans) feeding an 8-tier fuzzy entity resolver built for OCR noise, with a hard validation gate before anything reaches the merged dataset.
Reads the marketing numbers every morning, reports in Slack
Text version: GA4 · Meta · Google Ads · Kajabi → Reason → Write → Slack
An agent with MCP tools over the four APIs, running on a hardened Hetzner VPS (key-only SSH, dual firewall, fail2ban, loopback-only services) for EU data residency, delivering to Slack.
Cosmetic-dentistry previews from a patient photo
Text version: Upload → Queue → Inference → Poll → Preview
A three-service SaaS: Next.js front end, FastAPI job API, and a self-hosted Qwen Image Edit worker on RunPod L40S GPUs, with queued jobs and polling.
Revenue numbers that reconcile, every night
Text version: ShipHero → n8n → SQL Server → Stored procs → Fact tables → Power BI
ShipHero → SQL Server ingestion via n8n and GraphQL into staging tables, stored procedures building silver and fact tables, and a parity audit against the legacy production system.