experience | Ignacio Cuñado Barral

Research Experience

Research Assistant | September 2025 – Present | Delft, Netherlands

Researching Relational Deep Learning by benchmarking Tabular Transformers against state-of-the-art graph learning models (RDL, RelGT, ft-PNA, ft-FraudGT) and hybrid configurations
Investigating graph coarsening strategies to extend applicability of heterogeneous relational graphs for improved scalability
Working under the supervision of Prof. Kshitij Anan at the Data Intensive Systems Lab

Bachelor Thesis | April 2025 – June 2025 | Delft, Netherlands

Researched Graph Learning Methods on Tabular Data under supervision of Prof. Kshitij Anan
Investigated local and global attention mechanisms for Graph Transformers in Relational Deep Learning
Achieved state-of-the-art results on classification and regression benchmarks with custom relation-aware local attention

Research Project | October 2024 – February 2025 | Milan, Italy

Collaborated with Professor Guglielmi and team to develop Bayesian approaches to clustering nations based on educational performance using PISA dataset
Designed and coded semi-parametric approach and Nested Dirichlet Process (Rodriguez, Dunson, and Gelfand 2008)
Presented findings at the Italian Society of Statistics (SIS 2025) in Genoa

Part-time Software Engineer | February 2024 – Present | Heersmerk, Netherlands

Migrated payment providers for subscription-based tools handling over $10k weekly volume
Developed analytics and reporting features for subscription renewals, discounts, and handling with redesigned UX
Automated license-key generation and delivery, reducing wait time from 1 day to 10 seconds
Streamlined invoice generation, enabling users to access all available documents efficiently in one go
Migrated, updated, and refactored medical charity website with over 100,000 monthly visitors

Data Science Intern | July 2025 – September 2025 | Rotterdam, Netherlands

Built a web-based tool to visualize company hierarchies and emissions data inheritance for over 3 million companies
Researched and developed a novel attribution method incorporating trading data to measure decarbonisation of funds and mandates
Improved candidate selection pipeline, increasing privacy and accuracy by over 50% while lowering inference time

AI Software Engineer Intern | April 2024 – July 2024 | Delft, Netherlands

Designed a microservice with Django to extract information and categorize grant documents with a modular pipeline using Meta’s Llama framework
Leveraged vector databases and NER for semantic search and instantiated BERTopic model to categorize grants
Exhaustively evaluated the models by combining F1 score with text summarisation, achieving combined score of 0.77
Automated the search of academic papers and lexicographically ranked them based on relevance with transformers, then integrated key findings and arguments into proposal documents with LLMs

Languages: Python, Java, C++, Scala, HTML/CSS, JavaScript, SQL, PHP, Julia, Solidity

Technologies & Frameworks: NumPy, Pandas, PyTorch, Spark, TensorFlow, Turing, Bash, Git, Spring, React, Node, Laravel

Tools & Platforms: Ethereum Developer Bootcamp, IBM Advanced Data Science Bootcamp