Fellowships for PHD Students

Foundational Model Improvements:
Novel model architectures
Novel training algorithms and methodologies
Multi-modal (e.g., text, image, video, audio) understanding and generation
Multi-lingual understanding and generation
Algorithms and workflows for acquiring and curating high-quality and diverse datasets for training
Foundation Model Evaluation:
Creation of new benchmarks for assessing foundation model capabilities
Methodologies for robust evaluation of generative AI systems, including agents
Efficient Generative AI:
Efficient training and inference for cloud and on-device applications
Compute and memory efficient handling of multi-modal long/infinite context
Improving efficiency of diffusion models, including discrete diffusion
Reasoning:
Commonsense and domain-specific (e.g., math, coding) reasoning
Temporal and spatial reasoning
Reasoning for planning
Test-time scaling
Creative problem solving (learning out-of-the-box thinking techniques)
Knowledge Grounding:
Approaches to ground (multi-modal) generation on up-to-date world, domain-specific, enterprise, or personal knowledge
Memory-augmented generative AI systems
AI truthfulness and hallucination reduction
Agentic AI:
Creation of autonomous systems capable of performing tasks, making decisions, and interacting with their environments
Multi-Agent systems and agent orchestration framework improvements
Customization and continual improvement of agents post deployment
Internationalizing agentic systems
Responsible Generative AI:
Red teaming (e.g., advanced red teaming approaches, automated red teaming for multi-modal models)
Improvement of foundation model RAI performance (e.g., robustness to jailbreaking and membership influence attacks; watermarking approaches; deepfake detection)
Responsible agentic AI (e.g., robustness of multi-agent systems, adherence to guardrails)
Privacy-preserving Generative AI
Interpretable AI:
Understanding mechanisms behind emergent AI capabilities (e.g. mechanistic interpretation)
Communicating AI agents’ rationales and search process (meta-cognition) with human stakeholders
Small-size LLM:
LLM distillation
LLM customization for specific skills
Knowledge-graph aware LLM
Long-context, small-size LLM customized for reasoning capability
Applications of Generative AI:
Systems that leverage Generative AI for advancing science and technology in areas such as physics, mathematics, chemistry, biology, hardware design, materials science, engineering, economics, healthcare, climate
Security:
Low-cost computer vision techniques for object interactions for physical security
Near-real-time anomaly detection at scale (e.g., streaming petabytes per hour) to identify malicious activity in Linux cloud-based environments
Running distributed big data, AI/ML, and/or streaming models with minimal latency and cost using host, network, and audit telemetry sources. Such data engineering can be used for selecting interesting subsets of data including complex, multi-event sequences, improving matching and micro-summarization efficiencies for continuous log processing, and improving event summarization processing efficiency while maintaining accuracy
Use of AI/ML (e.g., GenAI) to automate and improve security response workflows (for example, from written reports or threat detections), specifically through incident summarization, response planning, and communication management

Topics

Eligibility

Important Deadlines

Application Instructions