IBM Ireland Graduate Fellows

Notre Dame PhD students in their third or fourth year studying Computer Science and Engineering, ACMS, Math, Physics, or related disciplines are eligible to conduct research at the IBM Research Lab in Dublin for 10 weeks over the summer from late May to early August each year.

IBM Ireland Graduate Fellow Award includes

• Round-trip airfare (up to a maximum of $1400, booked through Anthony Travel)
• Accommodation on Trinity College Dublin campus
• Living allowance ($4,500)
• GeoBlue health insurance
• Ground transport to and from work placement where necessary (TFI [Transport for Ireland] Leap Card allowance)
• Airport pick-up on arrival
• Professional development programming
• Cultural enrichment program

The IBM Ireland Graduate Fellows are administered through the Notre Dame Dublin Irish Internship Program.

Apply via Handshake

Applications are collected during the fall semester via the Notre Dame Dublin Handshake page under the IBM Ireland PhD Research Internships posting. The application package must include a resume, cover letter/statement of interest, and a letter of recommendation from your PhD supervisor.

Available research projects at IBM Ireland are detailed in the post on Handshake.

Successfully nominated Notre Dame PhD students will work as part of a research team at IBM Ireland. As such, they will experience life in an industrial research lab, working in a collaborative team environment, be responsible for their work, and develop new skills quickly and efficiently.

There is no single "ideal" intern profile. IBM Ireland expects interns to be generally curious, self-starting, and good communicators. The "typical" common skills of a successful intern include:

• Post-graduate experience in areas such as Computer Science, Engineering, Mathematics, or related area, or equivalent work experience in an industry setting.
• Previous AI/ML knowledge is essential. Most of the projects involve understanding how to apply AI/ML to solve real-world problems.
• Software development (Python) skills and statistics knowledge, to perform data preparation and analysis tasks relatively independently.

Depending on the individual project, additional useful skills may include:

• Familiarity with a deep learning framework (PyTorch, Keras, or Tensorflow)
• Applied knowledge of state-of-the-art LLMs and NLP
• Statistics, Machine Learning, Data Science
• Data Visualization, UI and UX Design
• AI foundation models, Graphical Neural Networks

Sample Research Projects from Summer 2025

Multi-modal Synthetic Data Generation

Our research team is seeking a highly skilled and motivated intern for the Spring/Summer of 2025. The ideal candidate will possess a strong foundation in AI and ML, with particular expertise in synthetic data generation, advancing current FMs, and integrating multi-modal data.

The internship will focus on advancing the generation capabilities of foundation models by integrating and analyzing multimodal data, including time-series, vision, and textual data. This integration will enhance the model's accuracy by incorporating domain-specific insights critical to understanding the system's dynamics. A crucial step in this framework is also ensuring that the generated data are of high quality.

Required Skills:

• A solid background in machine learning and optimization
• Familiarity with foundation models and their applications in data science
• Experience with multi-modality data, particularly time-series and spatial data
• Strong programming skills, especially in Python

Managing Knowledge Conflicts in LLMs

Knowledge conflicts are commonly presented to LLMs, and exploring the capability of the model to understand and manage them to ensure trustworthiness of the answer is gaining increasing interest in the community. There exist three categories of conflicts: intra-memory, context-memory, and inter-context. The student will focus on exploring the characteristics of these conflicts in a RAG approach. RAG has been proven to enhance LLMs’ capabilities in dealing with hallucination and enrich LLMs’ responses by integrating content from retrieved documents into the context. We aim to assess how well LLMs perform in dealing with real-world scenarios, rather than with synthetically created conflicts, to better understand their behaviour and capability.

Required Skills:

• Strong background in LLMs
• Strong programming skills in Python

Multimodal Foundation Models for Accelerating Scientific Discovery

This project aims to enhance Foundation Models (FMs) in complex scientific domains, particularly in drug discovery, by leveraging multimodal representation learning, existing Foundation Models, and LLMs.

The aim is to explore early fusion of multimodality for small molecules, combining diverse modalities and models to create a unified, compact representation. Given a small molecule, which can be represented as a SMILES, SELFIES, 2D graph, Image, FingerPrints, or 3D structure, we will design a multi-modal input multi-modal output model that aims at fusion of different modalities and utilises the multiview information to learn a representation that enhances prediction and generative tasks. By leveraging this multiview approach, the model aims to outperform late fusion and unimodal methods in capturing meaningful molecular representations.

Through this exploration, the project seeks to evaluate whether early fusion can provide a deeper understanding of small molecules, enabling better downstream applications in drug discovery.

Required Skills:

• Strong programming skills in Python
• Practical experience of deep learning frameworks (PyTorch, Keras) and research experience in representation learning - LLMs, GNNs, Molecule or Protein Sequences is a plus

Quantum Phase Transitions (SZ)

The proposed project aims to explore quantum phase transitions (QPTs), which are critical phenomena driven by quantum fluctuations at absolute zero temperature. The focus is on developing and utilizing quantum circuits to detect and characterize QPTs through the lens of quantum information theory. By leveraging quantum fidelity susceptibility - a measure of how sensitively a quantum state responds to small changes in system parameters - we aim to identify critical points and understand the underlying physics of phase transitions. The work will make use of advanced mathematical tools such as differential geometry, quantum information, partial differential equations, and Lie algebras/groups.

Required Skills:

• Tensor Networks
• Python
• Julia
• Quantum physics

Quantum-centric Supercomputing

The IBM Quantum team at the Dublin Research Lab is focused on developing utility-scale algorithms for Quantum-centric Supercomputing. The team has a strong focus on combining Tensor Network-based methods with quantum computing for near-term applications such as quantum simulation. This role will involve algorithm development as well as programming in Python and Julia.

Required Skills:

• Tensor Networks
• Python
• Julia
• Quantum physics

Contact Maggie Arriola Fagan, associate director for partnerships and engagment at Notre Dame Dublin, with questions.