About us

At IBM Research Europe in Dublin, Ireland, our open approach to research, coupled with the deep curiosity of our researchers, creates breakthrough client focused outcomes in domains such as IoT/Digital Twin, AI Security, Privacy, Healthcare, and Cloud. Together, we define and test new technologies on real business problems, discovering new growth opportunities and contributing to the success of IBM platforms and solutions.

Data Privacy


Data Privacy Framework to Manage Risks in Large Datasets

Our end-to-end framework for helping researchers and decision makers design and enforce a data privacy strategy for their use cases offers two major contributions compared to existing approaches. First, it provides a comprehensive workflow accompanied by a risk/utility exploration framework that supports informed decision making through detailed reporting and fast exploration of the anonymization space. Second, it is designed and implemented to scale with data, making the data privacy domain suited to the modern era of enormous and linked datasets.

Government Supported Collaborations


Learning at Scale


Feature engineering

Learning at Scale

We are working on next-generation data science services and acceleration of real-world data science projects by 10–100×. Our team’s foundations lie in data mining, machine learning, signal processing, and physics. We build systems that can handle huge amounts of spatiotemporal data, combined with geographic information and unstructured data such as text or images.

Current projects

AI for Physical Problems

Traditional physics-based models are driven by external forces: Tides rise and fall, winds blow in different directions, the depth and physical properties of water influence the speed and height of waves. These physical processes and their relationships are encapsulated in the differential equations that are coded into numerical models of wave transport. The nature of the computations typically demands high-performance computing infrastructure to resolve the equations. This high computational expense limits the spatial resolution, physical processes and time scales that can be investigated by a real-time forecasting platform.

We have developed a deep-learning framework that provides a 12,000% acceleration over these physics-based models at comparable levels of accuracy. The validated deep-learning framework can be used to perform real-time forecasts of wave conditions using available forecasts of boundary wave conditions, ocean currents, and winds. The huge reduction in computational expense means that

  • Simulations can be made on a laptop or tablet or programmable IoT device rather than a high-performance computing centre,
  • It enables investigation of a vastly increased set of physical conditions, geometries and time scales by amending input datasets to the deep-learning model.

IoT Data Curation Services

According to Forbes, data scientists spend more than 70% of their time collecting, cleaning and organizing data. For IoT applications, the effort is particularly high because of the volume, complexity and volatility of the data. To support data scientists in developing and maintaining IoT applications, we are developing data curation services that enable automated data exploration and self-healing data services, which can automatically detect, diagnose and resolve data inconsistencies.

Automated data exploration leverages machine-learning models to contextualize IoT sensor data, semantic models to capture domain-specific context and probabilistic data fusion techniques to ultimately derive a coherent view of the world from the heterogeneous, often conflicting data streams. A layer of self-healing data services is then derived, consisting of data processing flows, which can automatically answer user queries by providing cleaned and consistent data. Where gaps or conflicts in the data cannot be resolved automatically, data curation services enter into a dialogue with domain experts and update their semantic model of the domain.

Automated Feature Engineering

Another key step in data science projects is the transformation of raw data into features that can be used as input for machine-learning models. Often the raw data are stored across multiple tables in a relational database and need to be combined in various ways. This process is called feature engineering. Besides data curation, this is the most tedious and time-consuming data science task.

In our One Button Machine project, we are building a system that automates the feature engineering process at the “push of a button.” The One Button Machine traverses the graph defined by the entities (tables) and relations (primary/foreign keys) of a relational database. Along the graph traversal, the One Button Machine computes aggregate features that can be used as input for machine-learning models.

We have successfully applied the One Button Machine in various data science competitions, where it outperformed the majority of human teams and ranked among the top 25–36% of participants. In a client project with a social service provider from the U.S., the One Button Machine helped improve the accuracy on a complex classification task (involving a database with more than 20 tables) from 57% to 64%.

Security and Automation for AI


AI team

AI Security and Automation

Over the past few decades, machine learning—and in particular deep learning—has been phenomenally successful with regard to data-driven modelling, achieving close-to-human performance on a variety of cognitive tasks such as image classification or speech recognition. Of particular significance is the deployment of such models in critical applications such as healthcare, autonomous cars, and security. Such applications require answers to far-reaching questions involving trust, robustness, security and privacy of deep-learning models. Another focus of our research is the automation of neural network synthesis based on evolutionary algorithm and reinforcement learning.

Current projects

Adversarial Robustness Toolbox

Adversarial attacks pose a real threat to the deployment of AI systems in security-critical applications. Virtually undetectable alterations of images, video, speech and other data have been crafted to confuse AI systems. Such alterations can be crafted even if the attacker doesn’t have exact knowledge of the architecture of the DNN or access to its parameters. Even more worrisome, adversarial attacks can be launched in the physical world: Instead of manipulating the pixels of a digital image, adversaries could evade face recognition systems by wearing specially designed glasses, or defeat visual recognition systems in autonomous vehicles by placing stickers on traffic signs. Our team have built an open-source software library called the Adversarial Robustness Toolbox (ART) to support both researchers and developers in defending DNNs against adversarial attacks and thereby making AI systems more secure.


Recent developments have shown that AI is able to drive cars autonomously, beat the world champion in the challenging game of Go, and outperform humans in many other repetitive tasks. Hence, there is great interest in these technologies, particularly in deep learning. However, applying it to data is not trivial. Our team is working on the automation and democratisation of artificial intelligence (AI). We are developing AI tools that are capable of training predictive machine-learning models autonomously. This will enable anyone to employ the latest deep-learning methods to their data regardless of one’s prior knowledge of AI. We want to support our clients in finding the most suitable model for their AI problem by making our AI transparent to AI. Users will be able to supervise the search by making their own recommendations to the AI or by providing memory, inference or prediction time constraints. They will be supported by selecting and combining models.

IoT/Digital Twin


Cognitive IoT: Improving our understanding of the world

Our scientists are developing various “digital twin’ technologies, ranging from building a virtual platform to test complex IoT systems with live and simulated data to forming a knowledge graph for IoT that combines reasoning with machine learning to allow a system to analyze and understand data autonomously throughout the entire lifecycle of a connected asset.

Current projects

Virtual Testing Platform

To test complex IoT Systems, our researchers are using a virtual platform. This allows designers and developers of transportation services to investigate large-scale connected car services. They achieve this by merging simulations of large-scale automotive IoT deployments with proof-of-concept capabilities provided by real-world vehicles. This platform is helping automotive partners to design their services at scale while accelerating time to market.

Digital Twins

We are developing AI technologies to connect and understand IoT data in new ways. We’re combining machine learning with knowledge graph reasoning to enhance data extracted from an IoT network. We’re also adding layers of semantic meaning to create new insights within the network. This technology is called Digital Thread and is the core of each Digital Twin. It connects information along the lifecycle stages into a knowledge graph. This graph then allows informed decisions and process automation.

Next Generation Cloud and Systems


Data-Centric Computing and Cloud

Our researchers are addressing challenges that conventional Cloud and edge datacentres are facing, more so in light of patterns and requirements in computing created by trending cognitive workloads and use cases.

IBM Research ireland data

Current projects

Edge Computing for Cognitive IoT

Our researchers are working on realising the vision of offering general-purpose computing and analytics services at the “edge” of the cognitive Internet of Things value chain. By doing so, we foresee the creation of cognitive computing services whose sustained value is conditioned on the ability of near real-time analytical processing, while still having Cloud services play their vital role for aggregation, batch processing and further data monetisation.

For example, The potential of edge computing is nowhere more obvious than with video analytics. High-definition (1080p) video cameras are becoming commonplace. The requirement for real-time insights into such video streams is driving the use of AI techniques such as deep neural networks for tasks including classification, object detection and extraction, and anomaly detection.

Our team have developed “Semantic Cache”, an approach that combines the low latency of edge deployments with the near-infinite resources available in the cloud. We use the well-known technique of caching to mask latency by executing AI inference for a particular input (e.g. video frame) in the cloud and storing the results on the edge against a “fingerprint”, or a hash code, based on features extracted from the input.

New Cloud Architectures and Stacks

In today’s Cloud datacenters, physical systems comprise individual server units that contribute processing, memory, accelerators and storage resources. The challenge is for these resources to be more efficient, flexible and agile. We are developing a vertically integrated “datacenter in a box” prototype to showcase the superiority of disaggregation in terms of scalability, efficiency, reliability, performance and energy reduction.

We are also experimenting with a universal microserver architecture and software ecosystem that will address the challenges of performance, power consumption and reliability in both conventional Cloud and edge datacentres.

Data-Centric Cognitive Systems

Our researchers are working on system software for sustained efficiency and productivity of next-generation systems, as driven both from data-centric patterns exhibited by trending workloads and/or the need to grow to unprecedented levels of capability computing. We do so by applying findings and derived solutions in various industries that employ computational methods to respond to pressing environmental, social, commercial and scientific challenges, such as controlling the impact of oil spills, advanced manufacturing and understanding physical phenomena.

We are also focusing on making high-performance computing more consumable in several industries that could benefit tremendously from shifting from manual labour-intensive processes to robust and adaptive computational appliances. We are specifically focusing our cognitive computational appliances on the field of formulated product design to shorten R&D time and achieve technically superior and more sustainable formulations.

Selected techniques, methods and implementing software artefacts from our research on data-centric systems have been incorporated into state-of-the-art machines, such as JURON.

IBM Research Europe, Ireland data

Puzzle Solving Toolkit

Our team have also developed a “Puzzle Solving Toolkit” using a 3D camera and Computer Vision to assist the visually impaired.

AI Driven Interaction


AI Driven Interaction

The Interactive AI group tackles solutions and algorithms where the user influences the system and the system influences the user. Our work includes recommender systems, dialogue systems and decision support solutions with a focus on explainability and interaction leveraging a wide range of techniques from machine learning, frequent pattern mining, reasoning, AI heuristic search and AI planning.

IBM Research AI Driven Interaction

Current projects

AI for Sales

The AI for sales project tackles various decision points in the sales pipeline to help sellers make better decisions and disseminate expertise throughout the organisation. From helping identify which organisations might be high potential new clients to helping build the right team to fulfil a sales deal, the team is helping transforming sales with AI.

To identify high potential clients we harnesses the complex network of client organisations and the expertise of sellers and teams to assist in identifying clients, provide an enriched understanding of the client and products of interest along with a solution for building the appropriate sales team that can work together to drive the opportunity forward.

Our team have developed an Opportunity Team Builder using AI to support sellers in identifying required roles for the opportunity based on the products that the client is interested in, recommending the best people to fulfil these roles, and predicting a win probability based on the current team composition to guide users in team formation.

AI for Dialogue Systems

Dialogue systems have many applications, such as customer support or question answering. Typically they have been limited to shallow single turn interactions. However more advanced applications such as career coaching or planning a trip require a much more complex multi-turn dialogue. Current limitations of conversational systems have made it difficult to support applications that require customization and context dependent interactions. We leverage domain-independent AI planning to automatically create dialogue plans, customized to guide a dialogue towards achieving a given goal. This support much more complex, personalized multi-turn dialogues giving the user a much richer interaction.

Interactive Recommender Systems

Recommender Systems are used with increasing frequency in a wide variety of domains ranging from e-commerce, tourism, health and on-line learning. The interaction between users and recommendations currently tends to be limited to shallow feedback such as providing ratings or filtering. Engaging the user in dialogue with the system presents a unique opportunity to elicit preferences, allow the user to understand more about the recommended item and provide the user with feedback on what would help the recommender generate more useful recommendations. We have developed a generic framework that can be built on top of existing recommender algorithm to support interactive recommendations through dialogue. This novel interaction mechanism is currently under pilot in IBM Watson Career Coach to help users in finding career goals and planning how to get there.

AI Planning

Underpinning a number of our solutions is AI planning. There is no question that learning and reasoning are two key components of human intelligence. If learning is about remembering the past experience, planning is about exploring ahead. We explore ahead in order to understand what we need to do to achieve a given goal. AI planning is about teaching a machine to compute a plan. Domain-independent planning is an active research area focused on solvers that can address a diverse range of planning problems, which might be very different from each other. This is fundamentally different from most software out there, which specializes on one type of a problem.

Our team has developed a domain-independent planner with essential capabilities. It computes optimal plans and, in doing so, it guarantees that it will not run out of memory. It can reason about the uncertainty ahead, and compute plans with contingency options. In slightly more technical terms, our domain-independent planner can deal with problems with so-called nondeterministic actions.

Automated Reasoning over Policy Rules

Policy documents, such as health insurance policies, typically define conditions, rules and regulations in natural language, but it is often difficult to manually discover inconsistencies or rules that are conflicting with each other. Furthermore, it may also be quite tedious to manually check if claims are compliant or violate certain rules and regulations described in a policy.

Our researchers are developing novel AI algorithms and technologies to automatically extract essential knowledge from policy documents, predict and explain the risks associated with the claims being audited against the policies. Therefore, our technical contributions which facilitate effective transparency and explainability will help claim auditors to do their job better as well as policy owners to better understand the rules and regulations that govern their domains of activity.

Integrated Health, Social Care and Human Behaviour Change


IBM Research Healthcare Ireland

Integrated Health, Social Care and Human Behaviour Change

A deeper understanding of the complex interactions between chronic conditions is key to the future sustainability of healthcare systems. As people age, they tend to acquire a combination of conditions such as diabetes, congestive heart failure and chronic obstructive pulmonary disorder. Although some conditions, if detected and managed, may not be life-threatening on their own, a combination of these conditions is a major cause of decreased quality of life — and increased burden on already over-strained care systems. Models of multi-morbidities that can help monitor, manage and assist medical professionals to make informed decisions are largely non-existent in today’s healthcare systems. In the ProACT project, we are working with our partners in a pan-European effort to transform our ability to provide effective care for conditions that will affect our aging populations by investigating how to create end-to-end healthcare support for complex multi-morbidities.

Current projects

Using IoT, AI and Cloud Technologies to Advance Home-Based Integrated Care

IBM is exploring how context-based analytics can help address challenges in integrated care. We are working on tools to support decision making and care delivery by looking at the individual as a whole, rather than a collection of issues. Getting the full picture requires integrating information from disparate enterprise systems, publicly available data and data from the Internet of Things. This information is processed using IBM cognitive analytics and provides a consolidated view of the individual, key insights and most appropriate actions, ultimately creating a digitally integrated care management system. In turn, this system can be used to drive behavioural change and improve patient outcomes.

As part of an EU H2020-funded project called ProACT, our team at IBM Research Europe in Ireland are working with partners in academia and industry to find new ways to use IoT, AI and Cloud technologies to advance self-management capabilities and home-based integrated care for persons with multimorbidity (PwM). Our Health & Person-Centred Knowledge Systems team in Dublin are building a holistic model for PwMs using data on conditions, vitals, self-reports and behavioural assessments.

We have developed a Health and Wellness Profile Builder (HWProfile), which is being tested during the ProACT trials in Ireland and Belgium and is aimed at representing a PwM through several interconnected dimensions: demographics, medical factors, self-reports and behavioural factors. The state of the PwM is assessed through the sensors and self-report questionnaires taken through the ProACT CareApp. Daily questions are a valuable method to collect a wide variety of self-report information such as breathlessness scores for COPD and CHF, mood and anxiety levels or information on medication adherence.

Our IBM Research team has also developed InterACT, a Cloud-based platform within the framework of ProACT. InterACT, built on top of IBM Cloud, is a set of authenticated services to manage de-identified health data and coordinate collaboration among data providers, data analytics (like the above-mentioned HWProfile) and data consumer.

Behavioural Change

Our researchers are working on a project that will revolutionise the way evidence around behaviour change interventions (BCIs) is synthesised and applied to advance our understanding of human behaviours and behaviour change and has a potential impact on diverse areas beyond public health. It seeks answers to: “What works, how well, for whom, in what setting, for what behaviours — and why?” by developing and applying cognitive technologies to the rapidly growing corpus of human behaviour change research literature. This technology will assist researchers and practitioners in the field in addition to informing public policy decisions.

IBM Research Europe in Ireland, with partners in the Royal College of Surgeons in Ireland, The Mater University Hospital as part of University College Dublin and Deciphex Ltd are investigating biophysics-inspired AI technologies in the surgery and pathology stages of colorectal cancer.

The team at IBM Research Europe in Ireland are fusing the mathematics of photon diffusion with AI to provide real-time tissue classification, tumour boundary delineation and blood perfusion patterns to augment surgical and pathology decision making.

Visiting IBM Research Europe in Dublin, Ireland


IBM Dublin Technology Campus, Building 3, Damastown Industrial Estate, Mulhuddart Dublin 15, Ireland

Directions to IBM Research Europe in Dublin, Ireland:

Take Exit 6 off M50 (signposted Cavan / Blanchardstown). Take Exit of N3 for Clonee / Damastown Industrial Estate. Take right at top of slip road and proceed straight for approx 500m. Upon entry to IBM Technology Campus, IBM Research Europe, Ireland, Building 3 will be signposted on the right hand side of the campus, just past the campus roundabout.

Upon Arrival to Building 3 Main Entrance:

Building 3 is the building with blue trim on windows. Park and enter by the main entrance. Press buzzer and ask the security officer to release the door for you. Check-in at our reception by advising the security guard who you are visiting and/or event you are attending and he will provide a visitor security badge. 

By public transport

Dublin Bus 38B serves the campus from Dublin City Centre. Also Dublin Bus 38D (which has a limited service) is non-stop after O’Connell Street to the IBM Campus.
There is an Expruss Bus Route 870, see schedule.

Discover what IBM is disrupting today