Junior computer vision

Pay in the destination countries

Monthly, in US dollars. Each country is an Imora admission package — open one for that route's fees, language requirement and deadline.

What does the day look like?

A typical day runs from 09:00 to 18:00, totaling 40 hours per week. About 25% of your time goes into training, evaluating, and fine-tuning neural network models. Another 30% is spent on edge optimization and quantization, converting PyTorch models into TensorRT or ONNX Runtime engines and optimizing C++ code to keep inference latency below 30 milliseconds on embedded targets. Around 20% covers sensor calibration and configuring LiDAR or RGB-D depth cameras via ROS 2 topics, 15% is data engineering, and 10% involves CI/CD pipelines and hardware-in-the-loop debugging.

While outsiders imagine autonomous robot demos, 75% to 85% of your day is routine engineering: plumbing data pipelines, troubleshooting drivers, managing thermal throttling, and debugging C++ memory leaks. Shift work is rare, though occasional on-call incidents occur if an automated factory line halts.

Where does this lead?