17 Nov
|
Sony Research India
|
India
17 Nov
Sony Research India
India
Sony Research India (SRI) is looking for a Computer Vision Research Engineer / Scientist in Computer Vision to strengthen its work in generative AI, content analysis and advanced computer vision. The selected candidate will join a team of research scientists and engineers working on video generation and editing, object detection and content analysis, scene understanding, multimodal models, and responsible AI. The role offers the chance to do publishable research while seeing that work reach real products We are looking for a researcher with a strong record of first-author publications at top venues and deep expertise in one or more of generative models, object detection and video understanding, or advanced computer vision. Strong programming skills in Python and PyTorch are expected, along with experience training large models on multi-GPU clusters. We value candidates who can work independently, lead projects from idea to deployment, mentor junior researchers and collaborate across teams and countries. Qualification:
• PhD with 0–4 years of industry experience, or Master's degree with 3–6 years of relevant industry research experience in computer vision. Mandatory
Skills:
• Solid track record of first-author publications at top venues (CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR) or journals (TPAMI, TIP, IJCV)
• Deep understanding of generative models: Diffusion and score-based models, flow matching, VAEs and transformer-based architectures
• Hands-on experience implementing and training image or video generation models, object detection, segmentation, and tracking (e.g., YOLO-family, DETR-based, and transformer-based detectors)
• Experience with video understanding tasks such as action recognition, temporal localization, or video classification
• Solid foundation in multi-view geometry, camera models, projective geometry, calibration and optimization (e.g., bundle adjustment)
• Working knowledge of Vision-Language Models and multimodal learning
• Strong programming skills in Python with deep experience in PyTorch
• Experience training large models on multi-GPU setups
• Strong mathematical background in linear algebra, probability, statistics and optimization
• Ability to read, reproduce and extend recent research papers independently. Preferred
Skills:
• Experience with video foundation models, Diffusion Transformers, long-form or controllable video generation, and video editing
• Experience with large-scale video content analysis: scene and shot detection, highlight detection, summarization, tagging, or quality assessment
• Experience with detection frameworks and tools such as Detectron2, MMDetection, Ultralytics and video understanding codebases
• Experience building content analysis systems for media, broadcast or streaming platforms
• Experience with Hugging Face Diffusers, Transformers and open-source video generation codebases
• Experience with model optimization and deployment: distillation, quantization, CUDA, TensorRT, ONNX
• Experience with large-scale data curation and annotation pipelines
• Industry research experience and a record of moving research into products
• Experience mentoring interns or leading small research teams. Project Details: The Research Scientist will contribute to research projects in generative AI and 3D vision, with the goal of producing high-impact publications, patents and technology that can be deployed in Sony products. Projects will span the following areas:
• Generative video: large-scale video generation models using Diffusion Transformers, latent Diffusion and flow matching, with a focus on temporal consistency, long-form generation, and camera, motion and identity control.
• Video editing and personalization: instruction-based editing,
inpainting and outpainting, object replacement, style transfer, relighting, and efficient adaptation using LoRA, adapters and ControlNet-style conditioning.
• Object detection and recognition: real-time and open-vocabulary object detection, instance and panoptic segmentation, multi-object tracking, re-identification, and detection with vision foundation models (e.g., DETR-style, grounding and promptable segmentation models).
• Video content analysis: action recognition, temporal action localization, scene and shot boundary detection, highlight detection, video summarisation, content tagging and metadata generation, OCR, and image and video quality and aesthetic assessment for large-scale media archives.
• Multimodal understanding: video Vision-Language Models and multimodal LLMs for captioning, question answering, temporal and spatial grounding, audio-visual learning and spatial reasoning.
• Alignment and evaluation: preference optimization for generative models, dataset creation, and benchmarks and metrics for generative and core vision tasks.
• Efficiency and deployment: step distillation, consistency models, quantization, pruning, efficient attention, CUDA/TensorRT optimization and real-time inference on edge hardware.
• Responsible AI: content moderation, deepfake detection, watermarking and provenance for generated media, and bias and safety analysis. Projects will involve the full research cycle: literature review, method design, large-scale training on GPU clusters, benchmarking, publication and transfer to engineering teams. The researcher will also work with Sony research groups worldwide and with academic partners, and will guide interns and junior researchers on related projects. Sony Research India is committed to equal opportunity in all its employment practices, policies and procedures and to ensuring that no worker or potential worker will receive less favourable treatment due to any characteristic protected under applicable local laws.
📌 Research Engineer (India)
🏢 Sony Research India
📍 India