Skip to content
We're hiringCareers

Come solve the hard memory problem — one of the hardest open problems in agent memory.

Most of AI memory is the easy case: one assistant remembering one user. We work on the unsolved one — letting whole teams of agents share a single memory that's fast, governed, and impossible to poison. If that sounds like the most interesting problem in AI right now, we should talk.

The unsolved problem you'd work on
shared bucketLead agentResearcherCoderReviewerWeb searchDatastore
  • parent
  • subagent
  • peer
  • tool
Careers[1 / 6]

Why this work matters.

AI is becoming teams of agents, and the thing that's missing is a memory they can all trust. Get it right and agent teams finally work; get it wrong and they contradict each other, repeat work, and get fed bad data. We're building that missing piece — sub-200ms recall, zero model degradation, poisoning-resistant by design.

Read the full mission →
How we work[2 / 6]

Our values, in practice.

The day-to-day, honestly — behavior, not slogans.

  • Hard problems, small team

    You'll own real surface area, not a sliver of it.

  • Ship and measure

    Fast recall and honest benchmarks are the job — we prove claims, we don't assert them.

  • Trust is the product

    Provenance, audit, and safety aren't add-ons; they're how we think.

  • Plain words, deep tech

    We explain memory to a buyer in one sentence and to a researcher in a paper.

Open roles[3 / 6]

Real ownership, not a sliver.

Concrete ownership areas, tied to the actual stack — Neon/Postgres, Cloudflare, MCP. Roles below are placeholders to seed the pattern; we'll confirm before launch.

  • Founding Engineer — Retrieval & Storage

    Remote · EngFull-time
    What you'll own
    Own the sub-200ms hybrid retrieval path on Neon/Postgres + Cloudflare.
    What we look for
    Systems depth, latency obsession, and a bias for measuring before claiming.
    View role →
  • Founding Engineer — Governance & Safety

    Remote · EngFull-time
    What you'll own
    Build the promotion gate, conflict resolution, and poisoning resistance.
    What we look for
    Care for correctness and provenance; threat-model thinking; calm under ambiguity.
    View role →
  • Developer Experience Engineer

    Remote · DXFull-time
    What you'll own
    SDKs, MCP plugin, docs — make ULTRAMEMORY a joy to adopt.
    What we look for
    Empathy for developers, sharp writing, and taste for clean interfaces.
    View role →
  • Founding Designer / Frontend

    Remote · DesignFull-time
    What you'll own
    Own the ultramemory-frontend system and the signature governance viz.
    What we look for
    Systems-minded design, strong frontend craft, and a love of making complexity legible.
    View role →

Don't see your role? We always make room for exceptional people — introduce yourself →

How we operate[4 / 6]

Perks, kept honest.

The practical stuff. Unconfirmed perks are marked as placeholders until we lock them in.

  • Remote-first
  • [Competitive salary + meaningful equity placeholder]
  • [Health / benefits placeholder]
  • Real ownership
  • Learning budget
  • [PTO policy placeholder]

Items in [brackets] are placeholders — we'll confirm specifics before any offer; we won't invent compensation or counts.

We're hiring

If the hardest problem in AI memory excites you, let's talk.