Multimodal AI Engineer — Vision & On-Device Inference

Razer; Lille; Sur place; EUR 70 000 - 110 000; Plein temps

Résumé du poste

Razer is seeking a high‑performing AI software engineer to design and implement on‑device inference for in‑game events, CV and audio processing pipelines, and multimodal models. You will collaborate with a global team across the top games, optimize performance, and contribute to the SDK integration.

You should be proficient in C++, have experience with OpenCV, PyTorch or TensorFlow, and understand real‑time inference and model compression for deployment on edge devices.

Qualifications

• Proficient in C++ for high‑performance implementation.
• Solid understanding of ML fundamentals (CNNs, transformers, multimodal fusion).
• Experience with OpenCV, PyTorch, or TensorFlow for model integration.
• Background in computer vision or audio event classification.
• Familiarity with dataset management, labeling, and augmentation techniques.
• Experience with real‑time inference and model compression for deployment.

Responsabilités

• Build AI modules for CV/audio-based in-game event detection.
• Implement data preprocessing and feature extraction pipelines.
• Train, validate, and deploy models across top 100–3000 games.
• Design video decluttering pipeline for gameplay quality refinement.
• Collaborate with SDK developers for on-device inference optimization.
• Ensure low-latency, high-accuracy inference performance.

Connaissances

C++

ML fundamentals

OpenCV

PyTorch

TensorFlow

Computer vision

Audio event classification

Real-time inference

Model compression

Performance tuning

Outils

OpenCV

PyTorch

TensorFlow

Description du poste

Razer is seeking a high‑performing AI software engineer to design and implement on‑device inference for in‑game events, CV and audio processing pipelines, and multimodal models. You will collaborate with a global team across the top games, optimize performance, and contribute to the SDK integration.

You should be proficient in C++, have experience with OpenCV, PyTorch or TensorFlow, and understand real‑time inference and model compression for deployment on edge devices.