Multimodal AI Researcher
Core
Research and develop AI models that integrate text, images, audio, video, and other data modalities to understand and reason across multiple types of data.
Role type
Senior IC multimodal AI researcher
Builds
Scalable training and inference pipelines for multimodal AI applications including vision-language models and multimodal large language models
Domain
Artificial Intelligence / Machine Learning / Computer Vision / Natural Language Processing
Deliverable
production ML models
Required skills
Deep learning, transformer architectures, Vision Transformers, contrastive and representation learning, cross-modal attention, self-supervised learning, foundation models, Python, PyTorch or TensorFlow, distributed training, GPU optimization
Preferred skills
Vision-language models (CLIP, BLIP, LLaVA, Florence, Qwen-VL), multimodal large language models, retrieval-augmented generation, speech or audio models, image/video/document/medical imaging datasets, publications at leading AI conferences, open-source AI contributions, Kubernetes, MLOps, model serving, reinforcement learning, RLHF, synthetic data generation
Technologies
PyTorch, TensorFlow, Hugging Face Transformers, OpenCV, CUDA, DeepSpeed, Ray, NumPy, Pandas, Git, Docker, Linux, SQL, AWS, Azure, Google Cloud, Kubernetes, Pinecone, Milvus, Weaviate, FAISS
Responsibilities
Design, implement, and assess multimodal architectures; Develop and optimize models for image captioning, visual question answering, document understanding, speech-text understanding, video understanding, and multimodal reasoning; Fine-tune, evaluate, and benchmark foundation models with domain-specific datasets; Build scalable training and inference pipelines; Run experiments, analyze findings, and prepare technical reports or research papers
Seniority
Senior, hands-on IC
