Applied Computer Vision Engineer (Deep Learning)

We are building advanced AI-driven image analysis systems that extract structured geometric features from complex visual data. The role involves developing and deploying deep learning models for object detection, segmentation, and precise feature extraction from high-resolution images.

The ideal candidate will have strong experience in computer vision, deep learning model training, and image processing pipelines, with the ability to take models from research experimentation to production deployment.

Design and develop deep learning models for image understanding tasks such as:

• object detection
• semantic and instance segmentation
• feature extraction and geometric analysis

Build and train models using modern deep learning frameworks such as PyTorch.

Develop detection pipelines using architectures like YOLO or similar real-time detection models.

Implement segmentation pipelines using models such as SAM, Mask R-CNN, Detectron2, or equivalent frameworks.

Design and implement image preprocessing and augmentation strategies to improve model robustness.

Develop post-processing algorithms to extract precise coordinates and vectorized representations of visual features.

Create and maintain training pipelines, evaluation metrics, and dataset management workflows.

Work with annotation tools and create scalable labeling strategies for large image datasets.

Optimize model performance for GPU training and efficient inference.

Collaborate with engineering teams to integrate models into production environments.

• Strong programming skills in Python.
• Hands-on experience with PyTorch or similar deep learning frameworks.
• Experience training and fine-tuning object detection models (YOLO, Faster R-CNN, etc.).
• Experience with image segmentation models such as Mask R-CNN, SAM, or similar.
• Strong understanding of computer vision fundamentals.
• Experience with OpenCV and image processing techniques.
• Experience with data augmentation libraries such as Albumentations or torchvision.
• Experience working with annotation tools like CVAT, Label Studio, or Roboflow.
• Experience training models on GPU environments (CUDA).
• Familiarity with Docker and containerized ML environments.

• Experience building end-to-end computer vision pipelines.
• Experience with geometric feature extraction or curve detection in images.
• Experience with skeletonization, edge detection, or contour analysis.
• Experience with experiment tracking tools (Weights & Biases, MLflow).
• Experience deploying models using TorchServe, Triton, or similar inference frameworks.
• Strong GitHub portfolio demonstrating real computer vision projects.

• 35 years experience in Computer Vision / Deep Learning
• Candidates with strong personal or open-source projects are highly encouraged to apply.

Experience with:

• Detectron2
• segmentation pipelines
• model optimization and quantization
• large-scale image dataset handling

• Opportunity to work on cutting-edge applied computer vision problems
• Access to large real-world image datasets
• High-performance GPU training environments
• Ability to build models that move from research to real production systems

We are looking for engineers who enjoy:

• solving challenging visual perception problems
• building practical AI systems
• experimenting with new deep learning architectures
• taking ownership of model development and deployment

Question 1: - Describe a computer vision model you have personally trained from scratch (not just fine-tuned).

Explain:

• the dataset size and source
• how the data was annotated
• which model architecture you used
• what problems you faced during training
• how you improved the model performance

Question 2:- You are given 10,000 images but only 800 are labeled for a segmentation task.

Explain how you would train a model effectively using this dataset. Describe the strategy you would use to improve performance.

Question 3:- You have trained a PyTorch computer vision model that works well in experiments.

How would you deploy it so that a production system can process images in real time?

Explain the tools and architecture you would use.

You can also send the responses to the above questions with your contact details on hire@webassic.com