🇫🇷 Paris, France · 12h ago
AI / Agentic Systems Engineer
Rubicon
LinkedInjuniorEnglish-friendly
AI / Agentic Systems EngineerParis, France | Full-time | Deep Tech / DefenseWe are looking for an exceptional engineer to build advanced AI infrastructure and agentic systems at the core of our company.The role sits at the intersection of state-of-the-art generative AI, software engineering, model infrastructure, data science and secure computing.We are looking for someone capable of working across the full AI stack—from models and inference to agent architectures, software infrastructure, evaluation and security.The environment is security-sensitive and predominantly on-premise, with compute and authorized devices operating across a private network.You will have substantial ownership over the technical architecture and implementation.The roleYou will research, evaluate and deploy state-of-the-art approaches in generative AI and agentic systems.This is a deeply technical and hands-on role. You will work with frontier and open-weight models, develop AI-enabled software infrastructure, build robust agentic workflows, and create the systems required to evaluate, monitor and improve their performance.A core focus will be reliability and verification: developing architectures capable of testing outputs, challenging assumptions, independently reviewing work, identifying inconsistencies and escalating uncertainty when appropriate.You will work across models, inference, software infrastructure, data, networking, security and developer tooling.What we are looking forWe are primarily looking for an exceptional software engineer with deep practical knowledge of modern AI systems.You should be comfortable building complex systems from first principles rather than assembling demos from existing frameworks.Strong experience in several of the following areas is expected:Software engineering: exceptional Python skills; backend and systems architecture; APIs; asynchronous systems; distributed systems; testing; CI/CD; Git; production-quality engineering.LLMs & Generative AI: frontier and open-weight models; tool/function calling; structured generation; context engineering; model routing; inference optimization; quantization; fine-tuning where appropriate.Agentic systems: agent architectures; planning and execution loops; tool use; orchestration; delegation; human-in-the-loop systems.AI coding systems: state-of-the-art coding agents; repository-level reasoning; automated code generation and review; test generation; debugging; software-engineering benchmarks.Memory & knowledge systems: RAG; embeddings; vector search; knowledge graphs; context management and compression; retrieval; provenance.Evaluation: LLM and agent evaluations; automated testing; adversarial evaluation; independent reviewer/critic architectures; regression suites; observability and reliability measurement.Data science: strong numerical reasoning; Python scientific stack; statistics; data pipelines; experimentation; model evaluation and reproducibility.Model infrastructure: local inference; GPU deployment; model serving; batching and caching; containers; orchestration; heterogeneous compute.Security: secure-by-design architecture; authentication and authorization; secrets management; encryption; sandboxing; network isolation; auditability; supply-chain security.Networking & on-prem infrastructure: Linux; Docker/containers; private networking; VPN/tailnet architectures; service discovery; secure communication across multiple machines and compute nodes.Databases & storage: SQL; object storage; vector databases; caching; durable data architectures for AI systems.You do not need to be an expert in every area above. We care more about exceptional engineering ability, intellectual range, speed of learning and the ability to reason across the full system.Particularly relevant experienceWe would be especially interested in candidates who have worked deeply with autonomous coding agents, multi-agent systems, local LLM infrastructure, AI evaluation systems, secure AI infrastructure or advanced developer tooling.Experience operating open-weight models on local GPU infrastructure is valuable.Experience in security-sensitive, sovereign, air-gapped or on-premise environments is a strong plus.Experience in defense is not required.The Technical ChallengeThe frontier of generative AI is moving extraordinarily quickly. Models are becoming more capable, agents more autonomous, and the boundary between model capability and software infrastructure increasingly blurred.The challenge is to turn these rapidly evolving capabilities into reliable, secure and production-grade systems. That means solving hard problems across model selection and routing, inference, tool use, orchestration, context management, evaluation, verification, observability, security and infrastructure—while maintaining the flexibility to incorporate new models and approaches as the state of the art evolves.Sourced from LinkedIn. Relocantly aggregates public job postings; apply on the original site.