AI Research Intern – Time Series Foundation Models

We are looking for research interns interested in advancing foundation models for multimodal and sensor-rich environments. Our research explores how AI systems can learn generalizable representations from heterogeneous real-world data, including time series, audio, vision, and physical sensor signals, and transfer that knowledge across tasks and domains. As an intern, you will work alongside Bosch researchers on open research problems at the intersection of multimodal learning, foundation models, and real-world sensing. You will have access to challenging datasets and applications from areas such as mobility, manufacturing, robotics, and intelligent systems, with opportunities to turn research ideas into publications and working prototypes. ## What You'll Work On Your research may explore topics such as: - Foundation models for time-series and multimodal data - Causal analysis and evidence-based reasoning - Learning shared representations across images, time series, and other sensing modalities - Self-supervised learning from large-scale unlabeled sensor datasets - Multimodal fusion and cross-modal representation learning - Pre-training strategies for heterogeneous and asynchronous data - Generative modeling and synthetic data for sensor-rich applications - Adapting pretrained models to new sensors, tasks, and domains with limited labeled data - Robust learning under missing modalities, noisy measurements, and distribution shifts - Efficient multimodal models for real-world and edge deployment - New architectures and evaluation methods for multimodal foundation modelsYou will formulate research questions, develop and evaluate new methods, build experimental prototypes, and work with Bosch researchers and domain experts to understand how these approaches behave on real-world problems. Strong research results may lead to publications at leading ML, AI, or signal-processing venues. ## Basic Qualifications - Currently pursuing a Ph.D. in Computer Science, Electrical/Computer Engineering, Machine Learning, AI, or a related field - Research experience with time-series foundation models, multimodal models, or self-supervised learning - Hands-on experience with PyTorch or similar frameworks - Research or project experience with multimodal learning, time-series modeling, representation learning, foundation models, or signal processing - Comfortable designing experiments and analyzing model behavior rather than only applying existing models - Strong Python programming skills - Ability to communicate research ideas clearly and work collaboratively ## Preferred Qualifications - Experience working with real-world sensor and telemetry signals from radar, LiDAR, IMU, or other multivariate time series - Experience with Transformers, generative models, contrastive learning, or cross-modal learning - Experience adapting large pretrained models through fine-tuning, parameter-efficient methods, or transfer learning - Familiarity with robustness, domain adaptation, missing-modality learning, or uncertainty estimation - Publications at venues such as NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, ICASSP, AAAI, or related conferences - Open-source implementations or research projects demonstrating strong experimental skills