IBM PhD Fellowship Awards

Supporting research and innovation

We invite applications for the IBM PhD Fellowship Awards from students whose graduate research work aligns with IBM’s strategic directions in Semiconductor Technology, Quantum Computing, Artificial Intelligence (AI), Multi-Cloud Computing and Hybrid Cloud Platforms for AI, Security, and Responsible Computing. Current topics of interest within these areas include:

Semiconductor Technology

  • Two-Dimensional (2D) Materials

    Theoretical and experimental studies with atomistic simulations for discovering fundamentals and requirements for new channel materials being considered as contenders to Silicon (Si) for Complementary Metal-Oxide Semiconductor (CMOS) logic scaling aimed for future Angstrom-scale technology nodes. Growth, characterization, and processing of 2D transition metal dichalcogenides (TMD) materials for advanced CMOS logic applications.

  • Back-End-Of-Line (BEOL) Interconnects

    Experimental research in synthesis and characterization of novel conductor and liner materials, nanoscale interconnect design and fabrication, interconnect performance and reliability testing.

    Theoretical & computational research in high-throughput discovery and screening of topological conductors for post-Copper (Cu) interconnects, including first-principles, quantum transport and machine learning methods.

Quantum Computing

  • Quantum Algorithms Theory

    Study of new algorithms and tools for both near-term and long-term quantum computing. This includes new techniques for getting the most out of quantum systems exploiting error mitigation or error correction, compiler research for optimizing circuits on hardware, and algorithm development for addressing problems like simulating quantum systems or quantum machine learning for fault tolerance.

  • Quantum Algorithms Engineering

    Get the most out of hardware and software capabilities to scale existing quantum algorithms to regimes where classical computers struggle to perform brute-force computations. This includes identifying relevant computational problems that map efficiently to native hardware, efficient utilization of circuit transpilation, error suppression, error mitigation and quantum workflow optimization, and consistently benchmarking quantum hardware through relevant quantum computational problems.

Multi-Cloud Computing and Hybrid Cloud Platforms for AI

  • Building a Vibrant AI Hardware Ecosystem

    Objective benchmarking (including energy and carbon modeling and measurements), optimization, and adaptation of AI workloads on a diverse set of hardware options, including exploration of software portability layers, performance optimizations, and scalability for key AI workloads like training and inference.

  • Tools for AI Model and Application Building

    Advancing the performance and efficiency of model pretraining, tuning, and inference processes and tools. Exploring best practices for application building, deployment, and management, from application patterns like Retrieval-Augmented Generation (RAG), agents, and successors, to hybrid cloud infrastructure for large-scale and performant AI systems.

  • Expanding the Ecosystem of Open Models and Datasets

    Diverse model sizes and target applications, including multilingual, multimodal, and science models tackling broad societal goals like material science, drug discovery, climate change, etc., in addition to language generation. Aid AI model builders with trustworthy datasets, improved tools, and better efficiency.


  • Generative AI for Security

    Training and fine-tuning large language models that contain the knowledge and skills necessary to reason about security, generate security content, and drive agents that automate security tasks from threat hunting, detection engineering, and incident response. These capabilities can assist and automate security practitioners and automate many security tasks.

  • Security of Generative AI

    While GenAI has proven valuable in a wide range of use cases, the models and applications they drive increase the attack surface due to new threats: data and model supply chain risks (poisoning), prompt injection attacks, hallucinations and other errors, and over privileged agents provided with credentials to sensitive systems and resources.

  • Quantum-safe cryptography

    Designing advanced cryptographic algorithms that resist quantum attackers; working on the standardization and integration of new quantum-safe algorithms into the existing ecosystem of cryptography; crypto agility and the development of sound methods such as cryptographic combiners to migrate to a quantum-safe future.

Exploratory AI Science

  • Mathematical Theory and Analysis of AI Architectures and Algorithms

    Algorithmic theory as applied to AI (probabilistic and learning) systems, including especially complexity theory, linear and multi-linear algebra, optimization, and circuit complexity.

  • Reasoning and Planning

    Novel AI architectures that learn to reason and plan, and do so both in a data and an energy efficient manner. Models derived from such architectures should learn neural representations that lead to learning solutions which are demonstrably as performant, concise and interpretable as possible. They must have distinctly novel reasoning capabilities, e.g., reasoning about analogical relationships, or the ability to encode second order contextual relationships, or be able to carry out abductive reasoning. They must also be able to perform real-world, novel, reasoning tasks that have not been encountered in any data used for training.

Foundation Models (FMs) and Large Language Models (LLMs)

  • Agentic Workflows

    Building LLM based agents that orchestrate multi models for complex workflows, emergent behavior and learning of AI agents that collaborate with each other and/or with subject matter experts, capturing knowledge created by AI agents, develop reasoning and context-aware answering.

  • Multi-modal foundation models

    Building FMs with multiple data and/or domain modalities in scientific domains encompassing physics, chemistry, materials science, life sciences, health care or climate science; determine uncertainty in multi-modal FMs, development of fusion and alignment algorithms.

Trustworthy AI

  • Trustworthy and Safe AI

    Clarifying the landscape and taxonomy of risks and impacts of AI safety, e.g., bias, harmful speech and actions, etc., and general trustworthiness, e.g., fitness for purpose. Building tools, methods, and benchmarks for detecting and mitigating those risks and concerns.

  • AI Policy and Regulations

    Exploration of the technical implications of AI that influence government policies and regulations. This includes areas of trust and safety discussed above, as well as other social and employment impacts.

Responsible Computing

  • Research at the intersections between ethical, legal, social, historical, cultural, epistemological, and technical aspects of computation, including the datasets, technologies, processes, and infrastructures involved; implications of advanced and emerging technologies – e.g., cloud computing, AI & machine learning, quantum computing - or technologies that interact with these capabilities; creation of sociotechnical strategies that constructively challenge and mitigate potentially harmful outcomes and create tangible places for collaborative and critical computational praxis.