index

janitor has a huge long-tail content problem: characters, creators, genres, relationships, scenes, tags, memories, and user intent all collide. Search cannot be a keyword box. Discovery has to understand taste, freshness, safety, and narrative fit.

Why this role exists

The best character for a user may be new, niche, mislabeled, or buried under louder content. We need search and recommendation systems that help people find what they mean, help creators get matched with their audience, and keep the whole ecosystem healthier.

What you’ll do

• Build full-text, semantic, and hybrid search over characters, creators, tags, and story metadata.
• Own ranking features, retrieval quality, query understanding, suggestions, and personalization.
• Design pipelines for content understanding, embeddings, moderation signals, and freshness.
• Measure discovery with practical metrics: successful starts, returns, saves, hides, and creator outcomes.
• Work with product and infra to keep search fast, debuggable, and resilient at high query volume.

What strong looks like

• You have shipped search or recommendations that real users depended on.
• You know Elasticsearch or OpenSearch deeply, and you know where they stop being enough.
• You can explain ranking quality without hiding behind vague ML language.
• You are comfortable with queues, data pipelines, online/offline evaluation, and production debugging.
• You care about creator incentives, not just click optimization.

Nice to have

Experience with vector search, LTR, graph features, trust and safety ranking, or large consumer content catalogs.

How to apply

Email [email protected] with one search or recommendation system you improved and how you knew it worked.