Computer vision engineer

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?

The day looks like the work in NVIDIA vision postings: designing, training and deploying neural reconstruction and detection models, building offline perception pipelines that ingest terabytes of sensor data, developing computer vision algorithms for preprocessing and postprocessing, integrating deep learning algorithms into large-scale production pipelines, specifying data and labelling requirements with annotation teams, defining KPI frameworks and analysing failure modes on large real and synthetic datasets.

Named tools: Python and C++ on Linux, PyTorch for training and fine-tuning, CUDA and TensorRT for GPU acceleration, containers and microservices for deployment, source control with CI/CD, multi-camera and multi-sensor fusion stacks using camera, radar and lidar, and simulation tools including synthetic data generators. Senior postings expect multi-view geometry, camera calibration, structure-from-motion, SLAM, optical flow and stereo. Training clever models is 20% to 30% of the work; the rest is data engineering and labelling quality, pipeline plumbing, debugging, evaluation, optimisation for speed and memory, documentation and cross-team coordination. Hours are standard full-time office or hybrid, with no rostered night work except on-call rotations for production perception systems in autonomous vehicles and robotics, plus deadline overtime before releases.

Where does this lead?