Real-Time Vision Systems Engineer (Edge & Cloud)

Koireader Technologies; United States; À distance; USD 120 000 - 170 000; Plein temps

Avantages offerts par ce poste

Meritocracy

NVIDIA hardware access

Autonomy in experimentation

Cutting-edge Generative AI tools

Résumé du poste

Koireader Technologies seeks a Systems-First Computer Vision Engineer to push live video AI into production with real-time pipelines built on GStreamer/RTSP. The role focuses on ultra-low latency inference across Edge (NVIDIA Jetson) and Cloud environments, integrating models into robust streaming systems.

You will maintain a modular Python framework, optimize performance, and ensure fault tolerance even when cameras go offline or frames are dropped.

Qualifications

• Deep expertise in video streaming protocols (RTSP, WebRTC) and processing tools.
• Experience with OpenCV and image processing fundamentals; geometry and color spaces.

Responsabilités

• Architect streaming pipelines for real-time video ingestion with multiple RTSP streams.
• Integrate trained models into production pipelines ensuring stability and low latency.
• Latency optimization across Edge devices (NVIDIA Jetson) and Cloud servers.
• Build fault-tolerant systems that recover from camera outages or dropped frames.
• Maintain and evolve a modular, efficient Python codebase for the vision framework.
• Diagnose bottlenecks across CPU, GPU, and network and implement fixes.

Connaissances

Video streaming expertise

Core vision stack

Production Python

Docker deployment

Math & statistics

Outils

GStreamer

FFmpeg

WebRTC

TensorRT

DeepStream

Triton Inference Server

PyTorch

TensorFlow

Kubernetes

Description du poste

Koireader Technologies seeks a Systems-First Computer Vision Engineer to push live video AI into production with real-time pipelines built on GStreamer/RTSP. The role focuses on ultra-low latency inference across Edge (NVIDIA Jetson) and Cloud environments, integrating models into robust streaming systems.

You will maintain a modular Python framework, optimize performance, and ensure fault tolerance even when cameras go offline or frames are dropped.