Edge AI Engineer (human)

Neura Robotics; Gemeindeverwaltungsverband Metzingen, Riederich; Vor Ort; EUR 75.000 - 110.000; Vollzeit

Zusammenfassung

Neura Robotics seeks an Edge AI Engineer to turn cutting-edge models into on-device intelligence for our robots. You will work at the intersection of ML, embedded systems, and robotics to make models fast, lean, and reliable on the hardware we ship.

Responsibilities include deploying models at the edge, optimizing toolchains (ONNX, TensorRT, AIMET), and profiling runtimes to meet strict latency and power budgets across NVIDIA Jetson and similar platforms.

Qualifikationen

• Master's or PhD in Computer Science, Electrical Engineering, Embedded Systems, or a related field.
• 3+ years in ML or embedded AI engineering with production deployment on embedded accelerators.
• Fluency with embedded AI hardware platforms (e.g., NVIDIA Jetson, Qualcomm IQ-series).
• Solid knowledge of model optimization (quantization, pruning, distillation, architecture search).
• Fluency with model optimization and deployment toolchains (ONNX, TensorRT, AIMET).
• Strong Python and C++, PyTorch experience; comfortable with embedded Linux and low-level profiling.

Aufgaben

• Deploy edge models to production-grade performance on embedded hardware.
• Own the toolchain across model export and inference optimization frameworks.
• Profile and tune models to meet latency, power, and memory budgets.
• Collaborate with Research, Hardware, Software, and Product to ship on-device AI.
• Build benchmarking, on-device evaluation, and performance regression tests across targets.
• Debug accuracy drops after quantization and investigate latency spikes on real chips.

Kenntnisse

Python

C++

PyTorch

Embedded Linux

Ausbildung

Master's or PhD in CS/EE/Embedded Systems

Tools

ONNX

TensorRT

AIMET

Jobbeschreibung

Your mission & challenges

As Edge AI Engineer, you turn cutting-edge models into production-grade, on-device intelligence for our robots. You work hands-on at the intersection of machine learning, embedded systems, and robotics. Your focus is making models fast, lean, and reliable on the hardware we ship.

• Deploy at the edge: Take models from trained to deployed using quantization, pruning, distillation, and every trick it takes to make them fast and lean on embedded accelerator platforms (e.g. NVIDIA Jetson, Qualcomm IQ-series).

• Own the toolchain: Work across model export and inference optimization frameworks (e.g. ONNX, TensorRT, AIMET) and the SDKs that turn a model into a working robot behavior.

• Optimize to the metal: Profile, analyze, and tune models and runtimes to meet the latency, power, and memory budgets of each hardware target.

• Partner across functions: Work closely with Research, Hardware, Software, and Product to bring on-device AI from research to shipped product.

• Set the bar for quality: Build and maintain benchmarking, on-device evaluation, and performance regression testing across every hardware target we ship.

• Solve the hard problems: Debug an accuracy drop after quantization, chase down a latency spike, and tackle the problems that only show up on the real chip.

What we can look forward to

• Strong academic foundation: A Master's or PhD in Computer Science, Electrical Engineering, Embedded Systems, or a related field.

• Proven experience: 3+ years in ML or embedded AI engineering, with real production deployment on embedded accelerators, not just papers.

• Hardware fluency: Hands-on experience with embedded AI hardware platforms (e.g. NVIDIA Jetson, Qualcomm IQ-series, or comparable).

• Optimization expertise: Solid knowledge of model optimization from the model side to the metal: quantization, pruning, distillation, and architecture search.

• Toolchain expertise: Fluency with common model optimization and deployment toolchains (e.g. ONNX, TensorRT, AIMET, or equivalent vendor SDKs).

• Technical depth: Strong Python and C++, PyTorch experience, and comfort with embedded Linux and low-level profiling.

• The right mindset: A conviction that the only real test is the target chip, not the training cluster.

• Collaboration & communication: The ability to work independently, make sound calls under uncertainty, and speak fluently to researchers, hardware engineers, and product alike. Professional English required; German is a strong plus (B2 to C1).