About this role
Work Flexibility: Hybrid
We are seeking a Computer Vision/AI Engineer to research, architect, and deploy state-of-the-art vision models for high-throughput, low-latency computer vision pipelines. In this role, you will work on problems spanning real-time object detection, multi-class classification, tracking, and instance segmentation. You will bridge cutting-edge applied research with edge/cloud production engineering—experimenting with advanced CNN backbones, Vision Transformers (ViTs), and multimodal architectures while optimizing inference speed for real-world constraint.
Technical Responsibilities
- Evaluate, benchmark, and adapt frontier vision architectures (e.g., YOLO variants, R-CNN, ViTs, Transformers, MobileNet backbones).
- Design, train, and fine-tune computer vision models using PyTorch, TorchVision etc.
- Architect data preprocessing, augmentation and annotation workflows for imbalanced and noisy visual datasets.
- Optimize deep learning models for strict latency/throughput budgets using TensorRT, ONNX Runtime, DeepStream etc.
- Apply model compression techniques including post-training quantization (INT8/FP16) and knowledge distillation.
- Implement active learning loops and continuous monitoring systems to track data drift and edge-case model failures in production.
Minimum Qualification
- Bachelor's degree in software engineering/ Computer Science or related discipline with 5+ years of work experience or Master’s in relevant disciplines with 5+ years of experience or PhD degree in relevant disciplines with 2+ years of work experience.
Preferred Qualifications
-
5+ years hands-on experience designing and training computer vision models in production or applied research environments.
-
Proficiency in Python and PyTorch (or TensorFlow); solid foundation in OpenCV and NumPy.
-
Proficiency in end-to-end MLOps practices.
-
In-depth theoretical and practical grasp of
-
Classic & modern CNNs (ResNet, EfficientNet, ConvNeXt).
-
Vision Transformers & attention mechanisms (ViT, Swin, DETR, RT-DETR).
-
*Modern real-time detector architectures (YOLO, RT-DETR, SSD).
-
Excellent communication skills with the ability to work cross-team.
Good to have
Experience with video streaming protocols and pipelines (RTSP, WebRTC, GStreamer, FFmpeg, or NVIDIA DeepStream).**
Travel Percentage: None