Principal AI Engineer - barcelona
4 days ago
Barcelona
Principal AI Architect Multi-Agent Memory Orchestration Remote | Hybrid in Andorra | Hybrid in New Delhi THE OPPORTUNITY Memorae is not another AI wrapper. We are building the memory layer — the persistent, intelligent substrate that allows AI agents to remember, reason over time, and act on behalf of humans across their fragmented digital lives. We are a venture-backed company with strong revenue traction, an established product in market, and a growing user base across multiple channels. We are now preparing for our next funding milestone — and we need the architect who will own the most critical system in our stack: the Multi-Agent Memory Orchestrator. This is the role where you design the brain of the product — the system that decides what to remember, what to forget, when to surface knowledge, and how multiple AI agents coordinate through shared memory. If you've been waiting for a role where cognitive science meets distributed systems at production scale, this is it. WHAT YOU'LL OWN You will be the single technical authority over the memory orchestration layer, end to end: • Architecture. Design and build the multi-agent orchestration framework — routing, inter-agent communication, shared context management, and conflict resolution across autonomous AI agents., • Memory Systems. Own the hybrid retrieval system combining vector search, knowledge graphs, and temporal decay models grounded in cognitive science — deciding when to retrieve what type of memory and how to fuse results., • Product-to-System Translation. Translate product vision into production agent architectures that serve real users across multiple channels at scale., • Protocol Design. Build and evolve the integration layer enabling Memorae to act as the memory substrate for third-party AI agents and services., • Team Elevation. Elevate a strong software engineering team into AI-native development — bridging the gap between traditional engineering excellence and agent-based system paradigms. WHAT WE NEED FROM YOU Non-negotiable experience and capabilities: • You have shipped multi-agent systems in production — not prototypes, not demos. Real users, real scale, real failure modes., • You have deep, hands-on experience with both vector databases and knowledge graphs, and critically, with combining them in hybrid retrieval architectures., • You understand orchestration frameworks (LangGraph, CrewAI, AutoGen, or equivalent) well enough to know when not to use one., • You can design retrieval pipelines that blend semantic search, graph traversal, and temporal signals — the kind of system that models like Ebbinghaus forgetting curves and Zeigarnik effects demand., • You have practical experience with protocol-level integration work connecting AI agents to external services and data sources., • You communicate architecturally — you can draw a system on a whiteboard and make a room of senior engineers and investors understand why it matters. Highly valued: • Background in cognitive science, computational neuroscience, or memory systems research — even informally., • Experience scaling AI systems in B2C (consumer-grade latency, UX sensitivity, personalization at scale)., • Track record mentoring engineers and building AI-capable teams from strong traditional engineering foundations., • Contributions to open-source AI/ML infrastructure projects. WHY EQUITY AT MEMORAE IS DIFFERENT We believe the Principal AI Architect should have a meaningful ownership stake in what they build. This is not a token options package. This is a real ownership position in a company building category-defining infrastructure. • Early-Stage Entry — You join before the next major valuation event. Your equity is priced at current-stage economics with significant upside as the company scales., • Meaningful Grant — We are offering 0.25%–1.0% equity, vesting over 4 years with a 1-year cliff — calibrated to reflect that this role shapes the core intellectual property of the company., • Double-Trigger Acceleration — In the event of an acquisition, your unvested equity accelerates. We protect the people who build., • Revenue-Backed Traction — This is not a pre-revenue bet. The company has demonstrated product-market fit and meaningful revenue growth, which de-risks your equity position significantly., • Category-Defining Positioning — Memorae sits at the intersection of AI agents and memory infrastructure — a category that the industry's most important players are converging on. Early ownership in the company building the standard matters. COMPENSATION Base Salary: (adjusted for location and seniority) Equity: 0.25% – 1.0% with 4-year vesting, 1-year cliff, double-trigger acceleration Location: Fully remote or hybrid from Andorra (HQ) / New Delhi (engineering hub) THE TECHNICAL LANDSCAPE You won't inherit a blank canvas. The foundational architecture is established and documented — your mandate is to evolve it into a world-class orchestration system: • Vector Search: Semantic vector retrieval at scale, • Knowledge Graphs: Relational knowledge graph memory for structured reasoning, • Integration Layer: Protocol-based integration layer for third-party agent connectivity, • Channels: WhatsApp, Telegram, email, native app — real multi-channel complexity, • Memory Framework: A memory framework grounded in cognitive science — Tulving's memory taxonomy, Ebbinghaus temporal modeling, and Zeigarnik-informed task persistence, • Product Architecture: Dual-architecture product design separating long-term knowledge memory from actionable working memory, with intelligent features for deferred ambiguity resolution and unstructured input processing WHY MEMORAE Every transformative technology creates a crisis of organization. Writing created libraries. The printing press created catalogs. The internet created search engines. AI agents are creating an unprecedented fragmentation of context and memory across dozens of tools and services. Memorae is building the organizational layer the AI agent economy requires — not a feature inside someone else's platform, but the persistent memory substrate that every agent will need. You would be joining a company with early-stage ownership economics, growth-stage product traction, and a thesis that the largest AI labs in the world are validating every quarter. HOW TO APPLY Send a note and your background to the founding team directly. We don't do algorithmic screening rounds. The process is: • A conversation with the CEO about vision, architecture, and what you'd build in the first 90 days., • A deep technical session: bring a system you've designed and walk us through the hardest decisions you made., • A short take-home challenge (compensated) focused on hybrid retrieval design., • Offer within one week of final interview.
Page 3