Memory BridgeAbout this role
Summary: Founding ML Engineer role focused on architecting production-grade memory systems for large language models. Responsibilities include building retrieval engines, privacy-aware filtering, and adaptive memory updates using Python and modern AI frameworks.
At Memory Bridge, we’re tackling one of the hardest problems in AI: Large language models perfect memory while preserving user privacy.
We’re looking for an exceptional ML Engineer who sees beyond simple RAG tutorials and understands that building production-grade memory systems requires deep software engineering expertise first, ML knowledge second. You’ll architect and build the core retrieval engine that powers personalized AI experiences across conversations, privacy-aware filtering, and adaptive memory updates.
This is a rare opportunity to join our team as the founding ML engineer, where your code will directly shape how AI systems remember and learn. If you’re a strong programmer who has evolved into NLP/ML, we should talk.
About You
- The ideal candidate has evolved from senior software development into NLP/ML, bringing the architectural thinking and code craftsmanship needed to tackle complex memory persistence, multi-tier retrieval, and real-time personalization challenges that separate production systems from proof-of-concepts.
- You champion frequent deployments. You believe in shipping code regularly and improving through real-world feedback rather than seeking perfection.
- You solve problems creatively. You tackle challenges with a can-do attitude, recognizing when to build new features versus strengthening foundations.
Key Responsibilities
- Design and implement the core memory subsystem architecture, including storage tiers, retrieval pipelines, and privacy-aware filtering mechanisms
- Build evaluation frameworks to measure memory retrieval quality, relevance scoring, and augmentation effectiveness with clear metrics and benchmarks
- Implement sophisticated memory extraction and classification systems that identify, rank, and surface relevant user context from chat histories
- Research and adapt cutting-edge memory architectures from academic papers and open-source implementations (e.g., MemGPT, Reflexion, cognitive architectures)
- Architect the memory update pipeline that learns from user interactions, including relevance feedback, correction mechanisms, and adaptive privacy classification
- Design hybrid search strategies that combine semantic embeddings with structured metadata, graph relationships, and temporal patterns for optimal memory retrieval
- Collaborate with PM to define technical constraints and feasibility for memory system features
- Takes ownership of production reliability for AI features—comfortable being on-call for prompt failures, extraction issues, or integration breakdowns
- Understanding of AI system failure modes and fallback strategies—knows when to gracefully degrade vs. retry vs. escalate to human review
Requirements (Must-haves)
- 5+ years total experience, with at least 3 years in software engineering and 2 years building ML/NLP systems
- Hands-on experience building LLM-powered systems with a focus on context management, memory persistence, or personalization features
- Production experience with vector databases (Pinecone, Weaviate, Chroma) and hybrid search systems combining semantic and keyword search
- Proven experience shipping LLM-powered features to production, with an understanding of prompt engineering, context window management, and API reliability patterns
- Hands-on experience with LLM orchestration frameworks (LangChain, LangGraph, Mem0, MemGPT, Letta) for building complex AI workflows
- Demonstrated ability to implement hybrid search systems combining dense vectors, sparse indices, and metadata filtering
- Strong debugging skills for RAG pipelines, prompt reliability, extraction accuracy, and function calling/MCP integrations
- Experience with agentic architectures, including tool calling, MCP (Model Context Protocol), and function orchestration for database access and external integrations
- Experience with privacy-preserving ML techniques or building systems with strict data isolation requirements
- Expert Python programming, plus working knowledge of JavaScript/TypeScript for API integration and JSON schema design for LLM tool definitions
Nice-to-Have Skills
- You’re familiar with underlying technologies like transformer networks, attention mechanisms, and how they contribute to models’ abilities to generate coherent responses, function calls, and other language tasks
- Experience with graph databases or knowledge graphs for representing user memory relationships
- Familiarity with testing LLM applications—prompt regression tests, evaluation datasets, and A/B testing for AI features
- Experience designing LLM-friendly APIs with structured outputs, streaming responses, and token usage optimization
- Experience with experiment tracking and model versioning for memory/retrieval systems
- Experience with privacy regulations (GDPR/CCPA) and implementing data deletion, consent management, or audit trails in ML systems
- Background in recommendation systems or personalization engines
- Contributions to open-source LLM/RAG projects or memory-related research
- Master’s or PhD degree in physics, biology, CS, Electrical Engineering, etc., can be helpful but not required
Tech Stack & Tools
Core
- Python – Primary language for ML/AI development
- FastAPI – For building high-performance async APIs
- Git/GitHub – Version control and collaboration workflow
- GitHub Actions – Familiarity with CI/CD automation pipelines
- JSON – Standard for structured data exchange and API contracts
Databases
- Vector Databases – Pinecone, Weaviate, Chroma, Qdrant
- PostgreSQL – Primary relational database for structured data
- MongoDB – Document store for flexible, user-defined memory structures
AI/ML Frameworks
- LLM Orchestration – LangChain, LangGraph, LlamaIndex, Haystack
- Model Provider APIs – OpenAI, Anthropic, Cohere, and others
Nice to Have
- Elasticsearch – Full-text search and log-based memory indexing
- Docker – For local development and containerized deployment
How To Apply
At Memory Bridge, we believe that diverse perspectives and experiences make us better, which is why we have a non-standard application process designed to promote inclusion and equity. We’re looking for the best fit for each of our roles, regardless of the type of companies in your background, so we encourage you to apply even if your skills and experiences don’t exactly match the job description. All we ask is that you answer a few in-depth questions in our application that would typically be asked at the start of an interview process. This helps speed things up by letting us get to know you and your skillset a bit better right out of the gate. Please be sure to answer each question; the updated resume is essential.
Education is not a requirement for our roles; however, if you receive an offer, you will need to include your most recent educational experience as part of our background check process.
Memory Bridge is an equal-opportunity employer and we’re excited to work with talented and empathetic people of all identities. Memory Bridge does not discriminate based on someone’s identity in any aspect of hiring or employment, as required by law and in line with our commitment to Diversity, Inclusion, Belonging, and Equity. Our code of conduct provides a beacon for the kind of company we strive to be, and we celebrate our differences because those differences are what allow us to make a product that serves a global user base. Memory Bridge will consider all qualified applicants, including those with criminal histories, consistent with applicable laws.
The anticipated application window is 30 days from the date the job is posted, unless the number of applicants requires it to close sooner or later, or if the position is filled.
How We Work
We are a remote-native company, intentionally designed for deep work and focused execution. We hire the best minds, wherever they are, and empower them to solve some of the hardest problems in AI. Our culture thrives on asynchronous communication, radical transparency, and a commitment to sustainable performance—not burnout.
If you’re obsessed with solving the AI context problem and want to ship products that will redefine human-computer interaction, we want to hear from you.