Research Engineer – Generative AI and Computer Vision

Job Description

Role Overview

The Research Engineer – Generative AI and Computer Vision focuses on advancing state-of-the-art Generative AI and deep learning methods for real-world computer vision applications. The role combines applied research, experimentation, and engineering to develop robust AI solutions for image understanding, synthetic data generation, visual inspection, and continuous model monitoring.

The role will contribute to applied research and development of scalable AI capabilities spanning the full computer vision lifecycle—from data preparation and model development to deployment, monitoring, and continuous improvement. The work will address practical challenges such as limited or imbalanced training data, rare events, changing image conditions, data drift, and model-performance degradation in operation.

This role offers the opportunity to work at the intersection of advanced AI research and real-world deployment, translating novel Generative AI and computer vision methods into reliable and scalable solutions across a range of industrial and business applications.

Roles & Responsibilities :

Generative AI and Synthetic Data: Research, design, and implement Generative AI approaches for creating realistic and diverse synthetic image datasets. Explore diffusion models, generative adversarial networks, image-to-image generation, controllable generation, and related techniques to address limited data, rare defects, class imbalance, and long-tail inspection scenarios.

• Computer Vision Model Development: Develop and evaluate deep learning models for industrial image inspection tasks such as image classification, defect detection, anomaly detection, object detection, segmentation, and representation learning. Work with CNN-based, transformer-based, and multimodal architectures.
• Data-Efficient Learning: Investigate methods including transfer learning, self-supervised learning, few-shot learning, domain adaptation, active learning, and synthetic data augmentation to reduce annotation effort and improve model performance under limited-data conditions.
• Model Monitoring and Robustness: Develop methods for continuously monitoring deployed inspection models and image-data pipelines. This includes detecting data drift, changes in image conditions, model-performance degradation, confidence shifts, unusual defect patterns, and out-of-distribution inputs.
• Research and Experimentation: Formulate research hypotheses, design controlled experiments and ablation studies, establish appropriate evaluation benchmarks, and analyze results with scientific rigor. Compare state-of-the-art methods and identify approaches suitable for industrial deployment.
• Prototype and Product Integration: Build proof-of-concept solutions and research prototypes using Python and modern deep learning frameworks. Collaborate with software, cloud, MLOps, manufacturing, and domain teams to integrate validated methods into scalable AI platforms and applications.
• Scalable AI Lifecycle: Contribute to capabilities covering data preparation, model training, validation, deployment, monitoring, retraining, and continuous improvement across the complete computer vision model lifecycle.
• Technology Transfer and Scientific Impact: Translate research outcomes into scalable and production-ready AI components. Document technical findings, communicate results to technical and business stakeholders, and contribute to publications, invention disclosures, patents, reusable frameworks, and Bosch’s intellectual-property portfolio.
• Cross-Functional Collaboration: Work closely with global research, engineering, manufacturing, and product teams to understand inspection requirements and convert real-world challenges into well-defined AI research and development problems.

Qualifications

Educational qualification:

• Ph.D. or M.S. or M. Tech from top Indian or foreign institutes (IITs, IIITs, IISc etc.) in Computer Science, Artificial Intelligence, Machine Learning, Data Science, Electrical Engineering, Electronics and Communication Engineering, Robotics, Mechatronics, Applied Mathematics, or a closely related field.

Experience :

• At least 3 years of relevant professional or applied research experience in computer vision, deep learning, Generative AI, machine learning, or related areas.

Mandatory/requires Skills :

Strong understanding of machine learning, deep learning, and computer vision fundamentals.

• Hands-on experience developing computer vision solutions using techniques such as convolutional neural networks (CNNs) and other sequence models where applicable, transformers, autoencoders, and other modern neural-network architectures.

• Strong proficiency in Python and hands-on experience with PyTorch. Experience with supporting libraries such as NumPy, pandas, OpenCV, scikit-learn, and common image-processing libraries.

• Practical experience with transformer-based vision models, such as Vision Transformers, hierarchical vision transformers, detection transformers, or related architectures.

• Hands-on knowledge of Generative AI methods for image data, particularly diffusion models, generative adversarial networks, variational autoencoders, or image-to-image generation methods.

• Experience designing synthetic data-generation or advanced data-augmentation pipelines for training and evaluating computer vision models.

• Experience with one or more computer vision tasks such as image classification, object detection, semantic or instance segmentation, anomaly detection, defect detection, or visual similarity learning.

• Experience with the Hugging Face ecosystem (models, libraries, and SDKs) for rapid prototyping and evaluation of state-of-the-art AI models.

• Hands-on experience with model evaluation methodologies and metrics (such as Precision, Recall, F1-score, AUC, and ROC), including the rigorous interpretation of results to optimize model decision thresholds.

• Ability to formulate research questions, design rigorous experiments and ablation studies, establish evaluation metrics, benchmark alternative approaches, and interpret empirical results.

• Understanding of model generalization, domain shift, data drift, out-of-distribution detection, uncertainty estimation, and model-performance monitoring.

• Good software-engineering practices, including version control, modular implementation, code reviews, testing, technical documentation, and reproducible experimentation.

• Strong analytical, problem-solving, presentation, and communication skills.

• Ability to collaborate effectively with interdisciplinary and geographically distributed teams.

Preferred Skills :

Experience applying computer vision to industrial quality inspection, automated optical inspection, manufacturing, robotics, medical imaging, autonomous systems, or other safety- or quality-critical applications.

• Experience with synthetic defect generation, rare-event simulation, controllable image generation, domain randomization, or simulation-to-real transfer.

• Familiarity with vision-language models, multimodal foundation models, zero-shot or open-vocabulary vision methods, and parameter-efficient model adaptation (PEFT techniques such as LoRA, QLoRA etc.).

• Knowledge of self-supervised learning, contrastive learning, active learning, weakly supervised learning, few-shot learning, and continual learning.

• Experience with monitoring metrics and techniques such as input-data drift, feature drift, prediction drift, confidence calibration, outlier detection, defect-rate analysis, and alert generation.

• Familiarity with MLOps practices and tools for experiment tracking, model registries, automated training pipelines, containerization, deployment, observability, and model lifecycle management.

• Experience deploying machine learning models using cloud platforms and container technologies such as Docker and Kubernetes.

…