Senior ML Engineer — Memory Systems
Build the memory layer that makes Primordia different from every other AI system.
About the role
Most AI systems forget everything after a conversation ends. We don't. Mnemonic — our persistent memory subsystem — is what allows Primordia to compound intelligence over time across sessions, users, and deployments. As the engineer building it, you'll be working on one of the hardest open problems in applied AI: making memory that's fast, accurate, compressed, and useful.
This is not a RAG wrapper. This is a purpose-built cognitive memory architecture with episodic, semantic, and procedural layers. You'll own the design, implementation, and optimization of the system that every other Primordia product depends on.
What you'll do
- Design and implement the Mnemonic persistent memory subsystem across episodic, semantic, and procedural layers
- Build and optimize memory retrieval pipelines for low-latency operational contexts
- Develop compression and consolidation algorithms that maintain signal while reducing storage overhead
- Collaborate with the Aletheia (truth verification) team to ensure memory integrity across sessions
- Instrument and monitor memory performance in production deployments
- Contribute to architecture decisions across the full Primordia cognitive stack
What we need
- 5+ years of ML engineering experience with a track record of shipping production systems
- Strong experience with vector databases (Pinecone, Weaviate, Qdrant, or similar)
- Familiarity with graph neural networks and memory-augmented neural architectures
- Proficiency in Python; experience with PyTorch or JAX
- Experience designing systems that handle real-time data at scale
- Comfort working in a small team where you own significant surface area
Nice to have
- Experience with cognitive architectures or neuroscience-inspired AI systems
- Background in information retrieval or knowledge graphs
- Prior work on long-context or memory-augmented transformers
- Contributions to relevant open-source projects
We're a small team building something genuinely new. If persistent AI memory is the problem you want to spend your career on, email Mike directly.