About the role
The memory team gives models context that lasts months, not minutes. Long context windows help within a session but do not solve the problem: people return to the same work over weeks, and an agent needs to recall a decision, a preference or a commitment without rereading everything. The team builds retrieval, consolidation and update mechanisms that keep what matters and let go of what does not.
This is the capability that makes Clexerl Go feel like a colleague rather than a tool. An agent coordinating a team's schedule and customer relationships has to remember who agreed to what, notice when facts change and never surface something it should have forgotten. The team also works with Parent company security on what may be stored and for how long.
Engineers on this team own systems end to end, from storage and indexing to the model-side behaviour. Evaluations are built before features: a memory change ships only if it improves recall over simulated months of use without raising error or privacy risk.
What you’ll do
- Design memory architectures that combine long-context attention, retrieval over vector and keyword indices, and learned summarisation or consolidation.
- Build the write path: deciding what to store, how to compress it and how to reconcile new facts with contradicting old ones.
- Engineer low-latency retrieval services with sensible ranking, recency handling and freshness guarantees.
- Create long-horizon evaluations that simulate months of interaction, including updates, corrections and deliberate distractors.
- Study failure modes such as stale memories, false recall, over-personalisation and leakage between users, and design guards for each.
- Work with Post-Training to teach models when to consult memory and when to admit they do not remember.
What you’ll bring
- Solid software engineering experience building production retrieval, search or data systems with real latency and consistency requirements.
- Working knowledge of embedding models, approximate nearest-neighbour indices, reranking and hybrid retrieval.
- Understanding of long-context model behaviour, including positional extrapolation, lost-in-the-middle effects and KV-cache cost.
- Experience evaluating retrieval-augmented systems with realistic data rather than clean benchmarks.
- Care for privacy and data lifecycle: retention, deletion, access scoping and user control are design inputs for you.
- Fluency in Python and at least one systems language such as Rust, Go or C++.
Nice to have
- Research on memory-augmented neural networks, continual learning or knowledge editing.
- Experience with knowledge graphs or structured stores combined with language models.
- Background in recommender systems or personalisation under strict privacy constraints.
- Experience with CRDTs or event-sourced data models for syncing state.
Work you might take on
- 01
Build a consolidation job that merges months of conversation into compact, queryable facts and measure recall loss against the raw log.
- 02
Create a simulated six-month user history with contradictions and corrections to benchmark memory update behaviour.
- 03
Implement a hybrid retriever that weighs recency, relevance and source reliability, and tune it on real Clexerl Go failure cases.
- 04
Design a deletion pipeline that provably removes a user's information from indices, summaries and caches on request.
Compensation
$300K – $480K
The US base salary range for this role. Every offer also includes equity in Clexerl and, when it spins out, in the venture you help build. We tell you the reasoning behind your number.
Benefits
Health, fully covered
Medical, dental and vision for you and your dependants, premiums paid in full.
Two kinds of equity
Equity in Clexerl, plus a stake in the venture you help build when it spins out.
22 weeks parental leave
Fully paid, for every parent, however your family arrives.
Rest that sticks
Flexible time off with a four-week minimum we actually check on, plus a company-wide week in December.
Learning and compute
A yearly budget for courses, conferences and books, and compute credits for side projects.
401(k) with a 4% match
Matched from your first paycheque and vested immediately.
How we hire
Apply 5 minutes
Send links to what you have built and a few lines on why this role. No cover letter needed.
Intro call 30 minutes
A conversation with the hiring manager about the work, the team and what you want next.
Work sample 2–4 hours
A paid take-home or a pairing session on a real-shaped problem. You pick the format.
Studio day One day
Meet the team, go deep on your work and ours. Remote or in person.
Offer Within 5 days
A written offer with the full range, the equity and the reasoning behind both.
Opens an email with a short template. Links beat résumés.