Computer Vision Research Engineer

Senior Computer Vision Research Engineer – Deep Learning & 3D Vision

Experience: 5+ years

Role Overview

We are looking for a Senior Computer Vision Research Engineer with strong fundamentals in computer vision and deep learning, and extensive hands-on experience building, modifying, training, and optimizing computer vision models from scratch.

The ideal candidate should understand neural networks at the layer and architecture level, be able to modify existing architectures or design custom models, and have strong experience in high-accuracy pixel-level segmentation, transformers, 3D vision, depth data, and large-scale model training.

Responsibilities

· Design, develop, and train computer vision models from scratch for segmentation, detection, feature extraction, and related vision tasks.

· Understand and modify neural network architectures at the layer/block level, including convolution, normalization, activation, attention, skip connections, encoders, and decoders.

· Work extensively with segmentation architectures such as U-Net, U-Net++, SegFormer, Swin Transformer, and similar CNN/Transformer-based architectures.

· Develop highly accurate pixel-level segmentation solutions where prediction accuracy may directly affect millimeter- or centimeter-level measurements.

· Work with 3D vision, depth maps, point clouds, RGB-D/stereo data, camera calibration, and geometric computer vision.

· Build and maintain robust model training pipelines, including preprocessing, augmentation, loss functions, optimizers, learning-rate scheduling, hyperparameter tuning, validation, and error analysis.

· Work with large-scale datasets, including dataset cleaning, annotation quality analysis, class imbalance, sampling strategies, and data validation.

· Optimize models for inference speed, GPU memory, latency, and accuracy using techniques such as mixed precision, quantization, pruning, ONNX, and TensorRT.

· Develop and optimize inference pipelines for GPU, edge, and production environments.

· Maintain clear documentation of model architectures, datasets, training configurations, experiments, metrics, and model versions.

· Perform systematic model evaluation and failure analysis using appropriate metrics such as IoU, Dice, precision, recall, F1, mAP, and application-specific accuracy metrics.

· Research, evaluate, and implement new computer vision and deep learning approaches where they provide practical improvements.

Required Skills

· 5+ years of hands-on experience in Computer Vision / Deep Learning.

· Strong experience with PyTorch or similar deep learning frameworks.

· Proven experience building and modifying computer vision models from scratch, rather than only fine-tuning pretrained models.

· Strong understanding of CNN architectures, individual layers, feature maps, receptive fields, and encoder-decoder architectures.

· Strong experience with semantic/pixel-level segmentation.

· Hands-on experience with U-Net, U-Net++, SegFormer, Swin Transformer, Vision Transformers, or similar architectures.

· Good understanding of transformers, attention mechanisms, and modern vision architectures.

· Experience with 3D data, depth maps, point clouds, stereo/RGB-D data, or related spatial vision problems.

· Strong understanding of optimizers, loss functions, hyperparameter tuning, training strategies, and model convergence.

· Experience working with large datasets and production-scale training pipelines.

· Experience with model optimization and high-performance inference.

· Strong understanding of core computer vision concepts, image processing, camera geometry, and model evaluation.

Good to Have

1. Exposure to MLOps practices, including:

1. Model and dataset versioning

2. DVC or equivalent tools

3. Experiment tracking

4. Model parameter and configuration tracking

5. Training metric tracking

6. Model artifact management

7. Reproducible training pipelines

2. Experience with experiment tracking and model lifecycle tools such as MLflow, ClearML, Weights & Biases, or similar platforms.

3. Experience with high-precision visual measurement systems.

4. Experience with CUDA, TensorRT, ONNX, NVIDIA Jetson, or edge deployment.

5. Experience reproducing research papers or developing custom architectures/loss functions.

Education

B.Tech / M.Tech / MS / PhD in Computer Science, Computer Engineering, Electronics, IT, AI/ML, or a related engineering discipline from a reputed institute.

kshama.raj@blueocean.systems

+91 9833964181

Experience:

• Computer vision Research: 6 years (Preferred)

Work Location: In person