Heimdall

Heimdall is a fully AI-based marketing-automation platform: campaigns researched, planned, and executed with agentic multi-step workflows, and humans kept in the loop on the decisions that matter. I built the backend that makes those workflows dependable enough to run unattended.

Generation with guardrails

Marketing automation fails in a specific way — unsupervised generation at scale produces content nobody should publish. Heimdall's pipeline runs as inspectable stages: research, planning, content generation, and assisted execution, with the expensive, irreversible steps sitting behind a human-in-the-loop checkpoint.

The platform underneath

Heimdall runs on Django and FastAPI with PostgreSQL, Redis, and Celery for background processing. The agentic layer is orchestrated with LangGraph so each step is an explicit, observable node rather than a black box — the difference between a demo and something a team trusts with its outbound.