The job

Whyze runs on an engine: account research, warmth signals, clip selection, resonance scoring, reporting. Your job is building the agentic workflows that do the repetitive parts at scale, so the humans spend their hours on judgment.

What you'll own

• Agentic pipelines for account research, engager cross-referencing, and warmth detection
• AI-assisted clip selection and cut-down drafting that editors finish, not fight
• Resonance scoring and reporting automation feeding the monthly readouts
• Internal tooling that makes a small team ship like a large one
• The boundary lines: what automates, what never does, and the QC hooks in between

Required expertise

• LLM APIs and agent frameworks in production: orchestration, tool use and function calling, structured outputs, failure handling
• Python or TypeScript at a professional level, with the glue skills real automation needs (APIs, webhooks, queues, schedulers)
• Retrieval systems: embeddings, vector search, and when plain search beats both
• Evaluation and human-review design: measuring output quality, building review hooks instead of bypassing them
• Data discipline: client data isolation, PII handling, audit trails on anything that touches client accounts

You, probably

• Have shipped automation for content, marketing, or data workflows people rely on daily
• Design for human review instead of around it
• Get more excited by "the editor saved four hours" than by the architecture diagram