Robotics Middleware Engineer (human)

AI Data Engineer (Mensch) @ Neura Robotics; Freising; On-site; EUR 90,000 - 130,000; Full time

Job summary

Neura Robotics seeks an AI Data Engineer (Mensch) to own the middleware integration layer for its NEURA Core Robot Software. You will manage ROS 2 across platforms, port OROCOS RTT, and bridge Apex.OS with DDS ecosystems, ensuring deterministic, low-latency communication.

Responsibilities include coordinating cross-middleware interfaces, tuning QoS, and enabling zero-copy IPC for high-frequency data flows in a safety-conscious robotics environment.

Qualifications

• Strong expertise in C++ with RT patterns and deterministic behavior.

Responsibilities

• Own the middleware integration layer across ROS 2, OROCOS RTT, Apex.OS and DDS infrastructures.
• Lead integration of multiple middleware paradigms (ROS 2, OROCOS RTT, Apex.OS, DDS) ensuring cross-stack coherence.
• Tune QoS and network behavior at the wire level for reliability and determinism.
• Bridge Zenoh-driven systems with existing multi-middleware stacks and ensure stable integration.
• Champion zero-copy IPC approaches (Iceoryx, rmw_iceoryx2) for high-frequency data exchange.
• Coordinate state/command contracts across middleware surfaces to maintain ecosystem coherence.

Skills

C++ real-time

ROS 2

OROCOS RTT

Apex.OS

DDS (FastDDS/Cyclone/RTI)

Zenoh

Iceoryx

Zero-copy IPC

QoS tuning

Cross-middleware coherence

Tools

ROS 2

OROCOS RTT

Apex.OS

Iceoryx

Zenoh

Job description

Your Mission & Challenges

• You will own the middleware integration layer for the NEURA Core Robot Software department, spanning the full robotics middleware spectrum with applications ranging from real-time control to fleet messaging. Production and cognitive robotics increasingly need multiple middleware paradigms running side-by-side.

• ROS 2 stack: End-to-end responsibility for the ROS 2 layer across all robot platforms: sensor driver nodes, ros2_control YAML composition, DDS QoS tuning, Nav2 configuration, launch system architecture, and the diagnostic framework. Different release versions starting from Humble and Jazzy are maintained for both long-term and short-term migration cycles.

• OROCOS RTT integration: Ownership of the port-based hard-RT component layer behind the dynamics and whole-body control solver pipelines, where execution needs to stay deterministic down to sub-millisecond precision - including an application and service interface which serves as a hardware abstraction layer.

• Apex.OS / Apex.Middleware: This covers the full path from a standard ROS 2 development setup to a safety-ready Apex.OS deployment. This involves applying Apex.OS with its ISO 26262 ASIL-D certification for providing a deterministic, safety-certifiable ROS 2 runtime for any product line that needs to meet functional safety requirements.

• DDS infrastructure: Depending on platform or customer context, different variants and versions of FastDDS, Cyclone DDS, and RTI Connext need to be integrated, extended or bridged. The depth requires understanding how QoS choices play out at the RTPS wire level, how network address schemes and discovery methods apply in real network conditions, and where fragmentation and reliability semantics matter in practice.

• Eclipse Zenoh: The integration of new Zenoh-driven systems in local or wide area networks involves bridging to an existing multi-middleware stack, verifying requirements, and establishing a stable integration.

• Zero-copy IPC: For single host contexts, use zero-copy transport via Iceoryx2, rmw_iceoryx2 for ROS 2, or any of the other middleware's shared memory transport as a means to establish inter-process communication (IPC), such as for 1 kHz real-time data exchange or bridging different software domains, avoiding any communication overhead.

• Managing cross-middleware interface coherence: Robot state and command contracts need to behave consistently across all middleware surfaces the stack exposes. Keeping those contracts aligned as the system evolves is an ongoing responsibility. This also involves harmonizing the use of timestamped sensor streams, multi-modal observations, port assignments, and topic type schemes across stacks so both the end-to-end data plane and the ecosystem remain coherent.

What we can look forward to

• C++ depth with real-time-safe patterns - lock-free queues, zero-copy semantics, allocation discipline, RT-safe logging.

• Operational understanding of how middleware QoS settings behave at the wire-protocol level - not just configuration syntax. Comfortable reasoning about reliability, durability, history depth, fragmentation, discovery, and the failure modes each produces under load.

• Production experience integrating middleware into a real-time robot control loop with deterministic timing requirements.

• Production hands-on in AT LEAST ONE of the four core middleware paradigms (treated as parallel valid entry paths): (a) ROS 2 with ros2_control hardware interface authoring and lifecycle node design (Jazzy or Humble); (b) OROCOS RTT - component authoring, port-based composition, hard-RT deployment; (c) Apex.OS / Apex.Middleware - deterministic deployment, especially in automotive or industrial SIL contexts; (d) direct DDS implementation work - FastDDS, Cyclone DDS, or RTI Connext at the configuration-and-tuning level (not consumer-of-defaults), including QoS profile design for production deployments.

Nice to Have

• Hands-on across MORE THAN ONE of the four paradigms above - multi-middleware experience is the role's distinguishing competence, not a baseline requirement.

• Eclipse Zenoh for fleet-scale distributed messaging or edge scenarios.

• Iceoryx or alternative zero-copy IPC integration.

• Multi-middleware bridging patterns: ros1_bridge, OROCOS-ROS2 component bridging, or custom abstraction layers.

• DDS Security plug-ins (authentication, access control, cryptographic transformation) for SIL-grade deployments.

• Open-source contributions to any of the major middleware ecosystems (ROS 2 core, OROCOS Toolchain, Apex.OS, eProsima FastDDS, Eclipse Cyclone DDS, RTI Connext community).