Superads is building AI agents that help marketers make better creative decisions — not by surfacing more data, but by actually reasoning through strategy on their behalf. The harder problem underneath that ambition is one we haven't fully solved yet: how do you build a system that understands a brand deeply enough that an AI can make meaningful decisions on its behalf?
That's the problem you'll be working on.
Not in the sense of parsing brand documents and calling it done. In the sense of figuring out how to represent what a brand actually is — its identity, its audience, its creative principles, the judgment calls that a senior strategist who'd worked with them for years would make instinctively — in a way that an AI system can reason over reliably. It's a harder and more interesting problem than most companies are working on in this space, and we don't think it has a clean solution yet.
What you'll do
- Build the knowledge layer that sits underneath Superads' AI agents — the systems that give them genuine understanding of a brand rather than just access to its documents.
- Design how brand knowledge is represented, stored, and continuously updated: the structures that let an agent reason about a brand the way a senior strategist would, not just retrieve what's been written down.
- Develop agentic search capabilities — systems that let agents reason about what they don't know, identify knowledge gaps, and navigate information dynamically rather than relying on fixed retrieval pipelines.
- Evaluate and iterate on architectures for knowledge retrieval and reasoning, identifying where current approaches break down and driving more interesting solutions.
- Translate emerging AI research and patterns into production-ready systems that measurably improve output quality and reliability.
- Work closely with the product and AI engineering teams to ensure the knowledge layer directly improves what users experience — staying connected to product impact even when the work takes you deep into infrastructure.
What you'll need to succeed
- 8+ years in high-performance engineering environments — ideally at companies pushing the frontier of enterprise AI, legal AI, or knowledge-intensive agent systems.
- Proven experience building agentic workflows and RAG systems in production, with a focus on making AI output trustworthy and reliable — not just functional.
- Deep understanding of how language models represent and use knowledge, well beyond prompt engineering or standard retrieval patterns. You've thought seriously about memory architectures, reasoning under uncertainty, and dynamic retrieval.
- Experience with LangChain or LangGraph (or equivalent) for building multi-agent systems; strong Python skills, ideally with FastAPI.
- Hands-on experience with a major cloud data platform — Snowflake, BigQuery, or D