Parking traces processed in recent generative data workflows.
AI engineer working across production ML, research, and infrastructure
Machine learning, from research to working systems.
I work on production machine learning, research, and infrastructure. Recently, most of that work has been in autonomous systems and high-performance ML.
Observed MFU range on one recent generative pipeline.
Reduction in one internal height-map processing path.
Long-context target for differentiable Sinkhorn attention work.
2020—Now / Selected path
How I got here
Select a year to see what I was studying or working on, and how it led to the research questions I work on now.
01 / Foundation
2020—2024
Data science at University College Cork
Studied the mathematics and computing I would later use in machine learning. Alongside that, a customer-facing job taught me about how people use technical systems.
02 / Applied ML
Rotterdam · 2022
Forecasting at Agnicio
Benchmarked proprietary forecasts against LightGBM, XGBoost, and ARIMA baselines, then helped turn that work toward a SaaS product that was easier to access.
03 / Developer systems
Microsoft Research · 2023
Making it easier to get started
Built authentication mocks and configuration tooling, evaluated Copilot output, and won internal recognition for LLM-based onboarding and education concepts.
04 / Research engineering
Microsoft Research + UCD · 2024
From tooling into medical AI
Returned to Microsoft to improve CMS preview workflows, completed the B.Sc., and began an M.Sc. focused on AI for medicine and medical research.
05 / Production ML
Valeo Vision Systems · 2025
Deep learning for autonomous driving
Started working on generative data, JAX optimization, and orchestration for autonomous driving, including the infrastructure that keeps experiments running reliably.
06 / Current direction
Europe · Now
Making the algorithms practical
I now work on production ML and long-context attention research, including block-wise differentiable Sinkhorn attention and hardware-aware JAX kernels.
Select a milestone or use the arrow keys.
Profile
What I bring to a team
I like working on a research idea and on what it takes to run it. That can mean writing a kernel, building the pipeline around it, deciding how to schedule the work, and measuring and testing what comes out.
Most of my recent work has involved long-context attention kernels, computer vision, generative data, and large-scale processing. The scheduling and orchestration matter too, especially when other engineers need to be able to rely on the system.
Work History
Experience
Deep Learning Engineer
- Built a generative AI workflow over 10,000+ parking video traces using segmentation, masking, and inpainting. Brought utilization into the roughly 30-45% MFU range within the pipeline's constraints.
- Wrote hardware-agnostic JAX optimizations that reduced one trace processing path by 89%, from 45 minutes to 5, after earlier 7x speedups in 3D point-cloud rasterization.
- Designed Airflow and Kubernetes pipelines for extraction, processing, and dynamic task routing across cloud and on-prem resources.
- Investigated agentic CI/CD workflows and documentation generation to make a complex codebase easier to work with.
Software Engineering Intern
- Built state and API mocks that made it faster and cheaper to preview live changes driven by a CMS.
- Contributed tools that sped up engineering iteration in both production and testing environments.
Software Engineering Intern
- Built authentication mocks and a builder-pattern configuration framework that made onboarding and backend setup simpler.
- Led work evaluating Copilot output using NLP, semantic accuracy, and topic modeling techniques.
- Won internal hackathon recognition for LLM-based onboarding and M365 education concepts.
Data Science & Machine Learning Intern
- Benchmarked proprietary forecasting models against strong Kaggle baselines and developed LightGBM, XGBoost, and ARIMA alternatives.
- Helped move the forecasting work toward a SaaS product that was easier to access and scale.
Showroom Sales Specialist
- Worked in a team grossing roughly EUR 5M annually and introduced a chatbot to support online sales activity.
Selected Research
Projects worth highlighting
Block-wise differentiable Sinkhorn attention
Memory-efficient optimal transport attention with a custom backward pass, VMEM packing, and TPU-oriented JAX/Pallas kernels.
Research detailsDomain shift in deep RL agents
Built reactive exploration methods to quantify domain shift magnitude across OpenAI Gym environments.
Multi-omics analysis for hepatic liver cancer
Integrated genomic, transcriptomic, and proteomic signals to search for predictive biomarkers using deep learning and statistical modeling.
Education
Degrees
M.Sc. Artificial Intelligence for Medicine and Medical Research
University College Dublin
Second Class Honours, Grade 1 (2:1) · GPA: 3.66 / 4.20
First Class Honours threshold: 3.68 / 4.20
B.Sc. Data Science and Analytics
University College Cork
Second Class Honours, Grade 1 (2:1)