Unify is the institutional-context layer agents draw on: a traversable filesystem agents navigate like a code repo, with a .context/ directory at every level where agents write what they learn during execution. Your job is to make that context accurate, retrievable, and compounding.
What you will do
Build hybrid retrieval that grounds every agent query in the customer's own documents, traces, and corrections.
Design the sleep-time distillation pipelines that turn raw execution traces into structured context between tasks.
Own the layout and lifecycle of the context filesystem: what gets written, where it lives, when it is superseded, and how agents discover it.
Measure retrieval quality against eval outcomes, not proxy metrics — the test is whether the next run is better.
You will thrive in this role if you
Have shipped search, retrieval, or knowledge systems and have strong opinions about where embeddings help and where they lie.
Care about data lifecycle problems: provenance, staleness, and contradiction are the interesting parts, not edge cases.
Enjoy designing systems that agents — not just humans — are the primary users of.