ONHUMAN is an AI robotics company developing autonomous general-purpose humanoid robots. Our goal is to build embodied AI systems that can perceive, reason, and act in the real world. ONHUMAN is headquartered in Madrid, Spain, and this role requires 5 days/week in-office collaboration.

Team & Role Overview

Our ONVULC team is responsible for developing the core AI systems that power humanoid autonomy. We are looking for an AI Engineer, Agentic Systems to design and advance the high-level reasoning, long-horizon planning, and autonomous decision-making frameworks that enable robots to execute complex tasks in open-world environments.

This role focuses on developing new agentic architectures that bridge foundation models and physical execution—spanning task decomposition, tool use (APIs and physical skills), error recovery, and dynamic re-planning based on environmental feedback that directly impacts robot intelligence.

Responsibilities

• Agentic Architecture Design: Design and develop multi-agent frameworks and orchestration layers that translate open-ended natural language instructions into actionable robot execution plans.

• Long-Horizon Reasoning & Planning: Build models and systems capable of structured reasoning, task decomposition, and common-sense reasoning over extended timelines.

• Dynamic Re-Planning & Recovery: Advance closed-loop agentic behaviors, enabling the robot to autonomously perceive execution failures, reason about exceptions, and adapt its plan on the fly.

• Tool & Skill Integration: Improve agent capabilities in discovering, selecting, and sequencing discrete physical skills and low-level control APIs.

• End-to-End System Lifecycle: Work across the agent lifecycle, from initial orchestration research and prompting strategies to runtime optimization, low-latency edge deployment, and on-device fallback mechanisms.

• Cross-Functional Collaboration: Collaborate closely with modeling, pretraining, video, and RL teams to integrate high-level agentic reasoning with low-level Vision-Language-Action (VLA) models and whole-body control stacks.

• Experimental Rigor: Design evaluation frameworks, simulation stress tests, and profiling tools to measure agent intent alignment, safety guardrails, and decision-making robustness.

• Contribute to New Paradigms: Contribute to the development of new embodied agent paradigms (e.g., ReAct-based middleware, symbolic-neural planning) tailored for physical robotics.

Requirements

• Experience with Agentic AI: Experience designing, prompting, and training agentic systems, multi-agent frameworks (e.g., LangChain, AutoGen, or custom orchestration stacks), or tool-use architectures.

• Modern AI Foundations: Strong understanding of frontier Large Language Models (LLMs), Vision-Language Models (VLMs), and how they function as semantic reasoning engines.

• Performance through Innovation: Experience improving agent success rates and task accuracy through architectural innovation, search algorithms (e.g., tree search, beam search), or state-tracking methodologies.

• Technical Proficiency: Proficiency in Python and familiarity with deep learning frameworks (PyTorch) alongside agentic orchestration tooling.

• Experimental Rigor: Strong experimental rigor, high attention to logic branching, and the ability to systematically debug non-deterministic agent trajectories.

• Software Engineering Skills: Solid software engineering skills with the ability to build reliable, maintainable, and low-latency codebases ready for real-time applications.

• Autonomy & Ambiguity: Ability to operate independently and drive clean, structured engineering solutions into highly complex, open-ended cognitive problems.

Bonus Qualifications

• Embodied AI Middleware: Experience developing or working with robotic agent middleware (e.g., ROSA, RAI frameworks) or connecting LLM planners to ROS 2/physical hardware.

• Affordance & Grounding Background: Familiarity with grounding semantic plans into physical environments (e.g., affordance models, world models, or environment feedback loops).

• Frontier Lab Experience: Experience working on agent capabilities or frontier foundation models at companies such as OpenAI, Google DeepMind, Anthropic, Meta, or xAI.

• Distributed Systems & Edge AI: Experience deploying large model pipelines onto edge compute hardware under strict latency, memory, or power constraints.

• Publication Record: Publication record in top-tier AI or robotics conferences (e.g., NeurIPS, ICLR, ICRA, IROS) focusing on planning, reasoning, or agentic systems.

Compensation

The pay offered for this position may vary based on several individual factors, including job-related knowledge, skills, and experience. The total compensation package may also include additional components/benefits depending on the specific role. This information will be shared if an employment offer is extended.