Build AI memory.
From anywhere.
We're a small team working on a hard problem: making AI agents that actually remember. Not a chatbot company. Not a wrapper startup. The foundational memory infrastructure that every serious AI agent will need.
The problems are genuinely hard.
Cognitive memory architecture. 5-layer hybrid search with sub-50ms latency. Automatic knowledge graph extraction from unstructured text. Reinforcement learning that continuously improves retrieval quality. Memory decay curves modeled on Ebbinghaus forgetting. These aren't straightforward engineering problems.
We care about getting these right, not just shipping something that works at demo scale. If you want to work on infrastructure that other developers depend on daily and that has to actually work — this is the kind of work we do.
How we work
Not a list of values. Actual principles we make decisions against.
Ship early, learn fast
We'd rather have real user feedback on something rough than spend months polishing something nobody's tried. Iteration speed matters more than perfection at this stage. We ship, see what breaks, and fix it.
Developer first, always
We build tools we'd want to use ourselves. If the integration is confusing, the API is awkward, or the docs don't answer the real question — that's a product failure, not a documentation issue.
Remote by default
We don't have an office, and we're not planning to open one. The team works async-first across time zones. We communicate in writing, trust each other to manage their time, and meet when it's useful.
Small team, high ownership
Every person on the team owns significant chunks of the product. You'll make real decisions about architecture, API design, and product direction — not just implement tickets.
The technical surface you'd own
Memory architecture
3-tier cognitive model. Promotion and decay pipelines.
Hybrid search
Semantic + BM25 + graph + importance + recency. Sub-50ms.
Knowledge graph
Auto entity extraction. Relationship mapping. Graph queries.
RL quality pipeline
6 automatic checks. Reinforcement feedback loops.
API & SDKs
Python + TypeScript. OpenAI-compatible. WebSocket support.
Integrations
LangChain, CrewAI, LangGraph, MCP, n8n, Zapier.
Open Positions
We're not actively hiring right now, but we're always interested in hearing from people who'd be genuinely excited about this work.
If you're a strong engineer who cares about developer experience, distributed systems, or ML infrastructure — send us an email. No job description needed. Tell us what you'd want to work on and why.
Explore more