Droisys - Computer Vision Software Engineer

Description :

Key Responsibilities :

Model Development & Optimization :

- Implement deep learning models for object detection, instance segmentation, and classification in structured visual domains.

- Develop and optimize models for real-time inference on edge devices using tools like TensorRT, ONNX, NVIDIA Jetson, and CUDA.

- Apply advanced techniques for model compression, quantization, and pruning to meet performance constraints

System Architecture :

- Design and own the entire computer vision pipeline from camera capture to final inference output.

- Ensure low-latency performance, particularly on embedded or GPU-powered edge hardware.

- Benchmark and profile model performance using tools like NVIDIA Nsight, trtexec, or custom profilers.

Data Engineering & Image Processing :

- Drive dataset curation, augmentation, and labeling strategies for diverse object classes.

- Develop efficient preprocessing and postprocessing pipelines, including noise reduction, image enhancement, and perspective correction.

Deployment & Productionization :

- Integrate models into production environments with CI/CD and ML Ops pipelines.

- Ensure secure, efficient, and fault-tolerant deployment of models on embedded systems or cloud-edge hybrids.

Research & Innovation :

- Stay current with the latest trends in deep learning, vision transformers, diffusion models, and real-time inference.

- Evaluate research papers and contribute to internal innovation through POCs and rapid prototyping.

Required Skills & Experience :

- 3- 6 years of industry experience in computer vision and machine learning, with at least 3-5 years in an architect or technical leadership role.

- Expert in Python, with strong experience in OpenCV, PyTorch, and/or TensorFlow.

- Strong grasp of deep learning concepts: CNNs, attention mechanisms, image segmentation, small-object detection.

- In-depth knowledge of image processing, perspective warping, homography, and visual tracking.

- Hands-on experience with model optimization and deployment using TensorRT, ONNX, CUDA, and Jetson platforms.

- Proficient in tools like Docker, Git, and MLFlow or other model tracking/versioning frameworks.

- Solid foundation in linear algebra, calculus, probability, and statistics.

- Demonstrated experience in building CV systems for real-time applications, especially on edge hardware.

Preferred Qualifications :

- Prior experience developing perception systems for structured, rule-based environments (e.g., card games, tabletop detection, automated gameplay, etc.)

- Knowledge of hardware-aware training, mixed-precision inference, or vision transformers

- Experience with C++/CUDA kernels for custom ops or performance enhancement

- Published papers, patents, or open-source contributions in the CV/ML space