AI/ML Engineer specialized in Computer Vision for Embedded Systems (Edge AI)

Safran; Granada; Vor Ort; EUR 55.000 - 75.000; Vollzeit

Benefits dieser Stelle

Health insurance

Flexible work schedule

International environment in Granada,

Zusammenfassung

Safran in Granada, Spain, seeks an AI/ML Engineer specialized in Computer Vision for Embedded Systems (Edge AI) to develop compact, battery-powered optronic vision solutions. You will handle data strategy, dataset generation, training, optimization, evaluation, deployment, and validation on Qualcomm QCS/QCM platforms.

You will train and fine-tune object-detection models including YOLO, quantize with Qualcomm QNN, and build deployment pipelines using Qualcomm AI Hub, delivering robust edge

Qualifikationen

• Bachelor’s or Master’s degree in Computer Science, AI, Telecommunications, Electronics, Software Engineering, or equivalent.
• At least three years’ professional experience developing computer-vision, machine-learning, or deep-learning solutions.
• Hands-on PyTorch and/or TensorFlow experience in development, training, fine-tuning, and experimentation.
• Experience training and fine-tuning image-classification and object-detection models, including YOLO architectures.
• Dataset expertise: collection, annotation, augmentation, balancing, versioning, quality control, and train/validation/test splits.
• Experience evaluating accuracy and robustness with suitable metrics, analysing errors, and improving models/datasets.
• Experience optimising and quantising networks for edge deployment, including Qualcomm QNN.
• Experience with Qualcomm AI Hub or equivalent conversion, compilation, profiling, and deployment toolchains.
• Embedded Linux knowledge; practical AI deployment/debugging on Qualcomm QCS/QCM or comparable ARM-based edge platforms.
• Proficiency in Python, Git, Linux, systematic debugging; strong analytical, communication, and professional English skills.

Aufgaben

• Define data-collection strategies reflecting product use cases, environments, sensor configurations, and target classes.
• Generate, curate, annotate, version, and quality-check image/video datasets for classification and object detection.
• Train and fine-tune classification and detection models with PyTorch and TensorFlow, including YOLO detectors.
• Design training/validation experiments, tune hyperparameters, analyse failures, and improve generalisation.
• Evaluate accuracy, precision, recall, F1, mAP, latency, memory, power, and robustness.
• Optimise and quantise networks with Qualcomm QNN, balancing quality, compute, memory, latency, and power.
• Build and maintain Qualcomm AI Hub pipelines: conversion, compilation, profiling, validation, and release packaging.
• Deploy and integrate models on Qualcomm QCS/QCM; troubleshoot preprocessing, inference, post-processing, and application interfaces.
• Validate end-to-end vision performance on representative hardware, cameras, datasets, and operational scenarios.
• Collaborate with embedded software, hardware, systems, verification, cybersecurity, and product teams; document reproducible workflows and test evidence.

Kenntnisse

Computer Vision

Deep Learning

PyTorch

TensorFlow

Edge AI

Qualcomm QNN

Ausbildung

Bachelor’s or Master’s degree

Tools

Qualcomm AI Hub

QNN SDK

Hexagon DSP/HTP

OpenCV

ONNX

Jobbeschreibung

Our Technology:

Safran is a global leader in optronics, delivering advanced electro-optical systems for defense, aerospace, and security. Our portfolio features cutting-edge multispectral imaging technologies, including both infrared and visible-spectrum sensors integrated into airborne, naval, and land platforms for surveillance, reconnaissance, targeting, and navigation. Armed forces worldwide trust our solutions for their reliability, performance, and technological superiority.

We’re also a world leader in Resilient Positioning, Navigation and Timing (R-PNT), improving reliability, performance, and safety in critical applications and high-risk operations, even in GPS/GNSS-denied environments. We provide virtually fail-safe GNSS (Global Navigation Satellite Systems) and PNT (Positioning, Navigation and Timing) solutions for military and commercial applications worldwide.

Our industry-leading solutions combine ultra-accurate synchronization (sub-ns) and frequency distribution services. Built on enhanced White Rabbit PTP (Precision Time Protocol) synchronization technology, our products and technical solutions support a broad set of applications extending from timing distribution in financial networks to robust, reliable positioning.

We also specialize in control systems and diagnostics for particle accelerators and scientific facilities, providing innovative, dependable solutions with timing at their core.

The Position:

We are looking for an AI/ML Engineer specialized in Computer Vision for Embedded Systems (Edge AI) to develop vision for compact, battery-powered optronic products. The role covers data strategy, dataset generation, training, optimization, evaluation, deployment, and validation on Qualcomm QCS/QCM platforms.

Within the Optronics team, you will train and fine-tune classification and object-detection models, including YOLO, for efficient edge inference. You will quantize models with Qualcomm QNN, build deployment workflows using Qualcomm AI Hub, and integrate and validate models on embedded targets.

You combine strong computer vision and deep-learning expertise with practical embedded-systems knowledge. You use PyTorch and TensorFlow, design data-collection and annotation strategies, evaluate model quality and runtime performance, and collaborate with embedded software, hardware, systems, and product teams to deliver accurate, efficient, and maintainable AI capabilities for demanding field environments.

What you’ll be doing:

Define data-collection strategies reflecting product use cases, environments, sensor configurations, and target classes.

Generate, curate, annotate, version, and quality-check image/video datasets for classification and object detection.

Train and fine-tune classification and detection models with PyTorch and TensorFlow, including YOLO detectors.

Design training/validation experiments, tune hyperparameters, analyse failures, and improve generalisation.

Evaluate accuracy, precision, recall, F1, mAP, latency, memory, power, and robustness.

Optimise and quantise networks with Qualcomm QNN, balancing quality, compute, memory, latency, and power.

Build and maintain Qualcomm AI Hub pipelines: conversion, compilation, profiling, validation, and release packaging.

Deploy and integrate models on Qualcomm QCS/QCM; troubleshoot preprocessing, inference, post-processing, and application interfaces.

Validate end-to-end vision performance on representative hardware, cameras, datasets, and operational scenarios.

Collaborate with embedded software, hardware, systems, verification, cybersecurity, and product teams; document reproducible workflows and test evidence.

What you Bring to Safran:

Bachelor’s or master’s degree in Computer Science, Artificial Intelligence, Telecommunications, Electronics, Software Engineering, or equivalent.

At least three years’ professional experience developing computer-vision, machine-learning, or deep-learning solutions.

Hands-on PyTorch and/or TensorFlow experience in development, training, fine-tuning, and experimentation.

Experience training and fine-tuning image-classification and object-detection models, including YOLO architectures.

Dataset expertise: collection, annotation, augmentation, balancing, versioning, quality control, and train/validation/test splits.

Experience evaluating accuracy and robustness with suitable metrics, analysing errors, and improving models/datasets.

Experience optimising and quantising networks for edge deployment, including Qualcomm QNN.

Experience with Qualcomm AI Hub or equivalent conversion, compilation, profiling, and deployment toolchains.

Embedded Linux knowledge; practical AI deployment/debugging on Qualcomm QCS/QCM or comparable ARM-based edge platforms.

Proficiency in Python, Git, Linux, systematic debugging; strong analytical, communication, and professional English skills.

Desirable skills.

The company will positively consider knowledge and experience in:

Advanced object-detection architectures, recent YOLO variants, transfer learning, distillation, pruning, and quantisation-aware training.

Qualcomm QCS/QCM, AI Hub, QNN SDK, Hexagon DSP/HTP acceleration, and on-target performance profiling.

ONNX, TensorFlow Lite, OpenCV, model conversion, and heterogeneous CPU/GPU/DSP/NPU optimisation.

MLOps: experiment tracking, dataset/model versioning, reproducible training, continuous integration, automated evaluation, and model release management.

Synthetic data generation, active learning, hard-negative mining, domain adaptation, and techniques for limited or imbalanced datasets.

Camera systems, image processing, infrared/multispectral imagery, video pipelines, calibration, tracking, or sensor fusion.

Battery-powered edge products, defense, aerospace, optronics, wearables, or rugged, high-reliability applications.

The keys of our offer:

Be part of a leading and expanding company.

Health insurance.

Flexible work schedule.

International and dynamic environment in Granada, Spain.

High possibilities to grow inside the company.

Professional Career Plan.

Safran is an equal opportunity employer who does not discriminate based on race, color, citizenship, religion, gender, sexual orientation, national origin, age, disability or any protected category.