The Science Foundation Ireland (SFI) Centre for Research Training (CRT) in Foundations of Data Science is a large-scale collaborative initiative between the University of Limerick (UL), Maynooth University (MU), University College Dublin (UCD), a rich ecosystem of multinational and indigenous companies and an international network of collaborators; it is coordinated by Skillnet Ireland and Technology Ireland ICT Skillnet. The programme is led by co-Directors Professor James Gleeson (UL) , Professor Claire Gormley (UCD) and Professor David Malone (MU). The academic team includes vice-Directors Professor Norma Bargary (UL), Professor Nial Friel (UCD) and Associate Professor Catherine Hurley (MU), and a wide network of over 120 academic supervisors. .
Currently the CRT hosts 112 full time doctoral students distributed essentially equally across the three Higher Education Institutions (HEIs): UL, MU and UCD. From the four recruitment intakes to date 48 of the cohort self-identify as female and 49 as non-EU, (19 nationalities represented) demonstrating the CRT’s strong commitment to equality, diversity, and inclusion.
The CRT programme offers students a unique career pathway that will see them develop cutting-edge tools and technologies to provide insights to help shape industry, academia, and policy across the country. It not only helps mobilise the power of data nationally for industry across multiple sectors but will also assist in establishing Ireland as a leading data science research nation with a sustainable flow of expertise. All CRT students undertake a mandatory 12-week training placement with one of the CRT Enterprise Alliance (EA), a 16-strong body of industry and enterprise organisations, where they train with EA partners to solve an industry-relevant problem.
We caught up with Eva Ryan and Kaavya Rekanar to chat to them about their placement experience in Analog Devices International and Valeo Vision Systems.
Eva Ryan, CRT Student
Placement: Analog Devices International
I completed my 12-week training placement remotely as part of the PCT Test Engineering team in Analog Devices International (ADI). My project built on the work Sarah Murphy (UL, 2019 CRT cohort) completed during her CRT training placement with ADI, which focused on creating models to predict product failure during the testing process.
ADI is a global leader in the design and manufacture of semiconductor units used in a broad range of electronic equipment. These units undergo multiple stages of testing before release to ensure they meet quality standards, with the cost of each test contributing to the overall product cost. The primary motivation for this project was to access if machine learning (ML) could help to reduce costs. The goal was to build ML models to predict if a unit would fail at the final stage, using measures from earlier in the testing process as inputs. Identifying and eliminating failing units early using such models would avoid incurring the cost of later tests.
Guidance from my ADI supervisor and the wider team, along with Sarah’s work as a starting point, allowed me to hit the ground running with the project. I was impressed that despite the remote nature of my placement I had very regular contact with my ADI supervisor and other team members. I also had an opportunity to present my work to the team on numerous occasions.
The ML models I built were subject to a trade-off between capture rate (percentage of correctly predicted failures) and fallout rate (percentage of passes misclassified as failures). We carried out a cost analysis to demonstrate the potential cost benefit of the different models to ADI, having accounted for both capture and fallout. Some potential for cost savings and areas for further investigation were identified. The analysis also identified what tests are most important when modelling failure at the final test stage, which could inform future coverage reviews of what parameters are tested before the final test. Overall, the results provided proof of concept for utilising ML methods in the testing process. The project also highlighted a need for extensive data cleaning when analysing semiconductor testing data. One of my key project outputs was R code to identify and remove old results for retested devices from the datasets.
I really enjoyed all aspects of my training placement and found it hugely beneficial for my professional development. I gained invaluable experience working with real-world data, from data cleaning right through to interpreting and communicating modelling results in a business context. Working with the PCT Test team and regularly presenting my work also greatly contributed to the development of my soft skills. I am very grateful to my supervisor Stephania Kigadye and everyone at ADI for the continuous guidance, mentoring and support I received throughout my training placement.
Kaavya Rekanar, CRT Student
Placement: Valeo Vision Systems
I did my 12-week summer placement during the summer of 2022 in a hybrid setting as part of the Research and Development team at Valeo, Tuam, which proved to be one of the most enriching experiences in my career thus far. My project was a novel idea originated from discussions surrounding the potential of integrating Computer Vision with Natural Language Processing, resulting in a ground-breaking concept. The project aimed to explore AI Explainability in the context of self-driving cars. For instance, if the car slows down at a certain point and the driver asks, “why did the car slow down?”, it (the car/system) should be able to answer with the reason in human-understandable language. While it is closely related to my PhD, it was great that I was placed in a team which shared the same enthusiasm for research as I did.
Regular meetings with my mentors at Valeo provided me with a comprehensive understanding of the industry’s perception of research, emphasizing that good research entails addressing industry-related problems. My project’s primary objective was to investigate whether existing deep learning models could integrate computer vision and natural language processing in autonomous driving. To achieve this goal, we decided to use transformers architectural models as they are incredibly fast, allowing them to be trained with huge amounts of data. We decided that it was best to train the model particularly on a dataset that is related to cars and driving because in our research, we found that there is no existing model that already does this at all.
We experimented with models like ViLBERT and VisualBERT, utilizing a meticulously curated dataset comprising relevant self-driving scenarios. Although the project encountered several challenges concerning the dataset, models, and experimentation, the project’s triumph would not have been possible without Valeo’s exceptionally proficient research team.
The model I developed successfully addressed queries regarding a driving scenario. My three-month summer training placement presented new research prospects, which I intend to explore further during my doctoral studies. During my placement, I delivered multiple presentations on my project, which proved to be an invaluable opportunity to receive feedback from diverse perspectives. In addition, it helped me refine my soft skills and social skills, making presentations less daunting for me. Collaborating with the Valeo research community enabled me to feel supported and motivated, leading to exceptional project outcomes. I am extremely grateful of the assistance and encouragement extended to me, and I regard Valeo’s dedication to exceptional research as truly empowering.
For further information please visit: Centre for Research Training (CRT) in Foundations of Data Science
