Role

You’ll own the AI product experience for Pocket, focusing on prompts, agent behaviors, evaluation, and reliability to deliver best-in-class summaries and assistant workflows.

You’ll work closely with product, design, and backend to turn messy real-world inputs into consistently great outputs. You’ll join a small, high-output team that ships at the level of world-class consumer tech companies.

Compensation: Competitive salary + equity

Location: San Francisco

Responsibilities

AI Product Engineering

• Design and iterate on prompt and agent systems that power summaries, structured notes, and “ask the assistant” experiences.
• Build robust tool-using agent workflows: retrieval, function calling, planning, and safe execution.
• Define output schemas and consistency rules so results are reliable, scannable, and user-trustworthy.
• Develop evaluation harnesses for quality: golden sets, regression tests, rubric-based scoring, and human review loops.
• Reduce failure modes: hallucinations, instruction drift, bad formatting, missing context, and unsafe outputs.
• Tune for latency and cost while preserving quality; implement caching and fallbacks.
• Partner with backend to productionize pipelines (queueing, retries, idempotency, observability).

Data & Feedback Loops

• Build feedback mechanisms and labeling workflows to continuously improve quality.
• Analyze failures, cluster issues, and translate them into prompt/tooling changes.
• Create dashboards/metrics for quality, latency, and user trust.