Second-Year Computer Science Student at University of Exeter
Open Source Contributor (Keras, FFmpeg) | Incoming Visa / Featurespace Intern | ML Researcher
Marcos Ashton Iglesias
Developer | Open Source Contributor | ML Researcher
Writing
About Me
I am a second-year Computer Science student at the University of Exeter with a first-class average. Born in Madrid, I am a Spanish citizen studying in the UK. My work spans open source contributions, machine learning research, and full-stack development.
I have 32 merged PRs to Keras (Google), implementing new ops and RNN backends across PyTorch and JAX, plus 8 merged patches to FFmpeg covering FATE test coverage and correctness fixes in libavutil and libavfilter. I am an incoming software engineering intern at Featurespace, a Visa company that builds real-time payments fraud and financial-crime prevention.
Technical Skills
Programming Languages
- Python
- C
- C++
- Rust
- TypeScript
- Go
- JavaScript
- Java
- HTML/CSS
Frameworks & Tools
- Keras
- TensorFlow
- PyTorch
- JAX
- Next.js
- React
- Chakra UI
- Scikit-learn
- Docker
- AWS
- Git
Hardware & Robotics
- Jetson Nano
- LIDAR
- ROS
- Sensor Fusion
Languages
- English (Fluent)
- Spanish (Native)
Education
University of Exeter
BSc Computer Science | 2024 - Present | First-Class Average
- Docker Certified
- Currently Learning Rust
Experience
Visa (Featurespace) — Software Engineering Intern
Incoming, 2026 | Real-time payments fraud and financial-crime prevention
Incoming intern at Featurespace, a Visa company that builds adaptive behavioural analytics for real-time fraud detection used by major banks and payment networks.
Open Source
Keras (Google) — 32 Merged PRs
cuDNN GRU/LSTM for JAX & PyTorch, new ops (matrix_rank, pinv, geomspace, fold), CTC beam search, ONNX export, performance, correctness fixes
FFmpeg — 8 Merged PRs
FATE test coverage for libavutil components and decoders, correctness fixes in libavfilter and libavutil
keras-team/keras
32 mergedAcross the JAX and PyTorch backends: cuDNN-backed RNN paths, new public ops, ONNX export work, performance optimisations, and a long tail of correctness and shape-handling fixes.
github.com/keras-team/kerasFFmpeg/FFmpeg
8 mergedFATE regression tests for libavutil components and decoders (film_grain_params, stereo3d, pcm_bluray, and more), plus correctness fixes in libavfilter and libavutil.
code.ffmpeg.org/FFmpeg/FFmpegTimeline
Hackathons & Achievements
Movember Hackathon
Built a skin cancer image classifier achieving 80% accuracy
Nebrija Hacking Competition
Placed 10th out of 150 competitors
Projects
Featured
Open Source
Experience
Interests
Machine Learning, Open Source, Robotics, Systems Programming, Rust, Computer Architecture, Mathematics, and Basketball.
Connect With Me
Get In Touch
I am open to opportunities in software development, open source, and machine learning.
Phone: +34 655 59 59 79