Senior Data Scientist building production-grade data pipelines, ML systems, and large-scale analytical workflows. Focused on distributed systems, cloud infrastructure, and real-world optimisation problems.
Open to remote opportunities in data infrastructure, ML systems, and high-performance data processing.
I design and build data and machine learning systems at scale. My work focuses on distributed data pipelines, real-time architectures, and applied ML, with production experience across cloud environments (GCP), event-driven systems (Kafka), and high-performance data processing (Spark, Polars). I began my career in academia as a computational biologist, publishing in journals such as Nature Communications and PNAS, where I developed deep learning models and large-scale simulations. Today, I bring the same rigor to industry problems, with an emphasis on scalability, performance, and measurable impact. Currently Data Scientist at Leroy Merlin, working on supply chain optimisation, pricing, and logistics systems.
A selection of systems and experiments focused on scalable data processing, machine learning, and simulation.
Open data analysis · 2026
End-to-end data pipeline processing ~3M records from Wikidata, combining high-performance data processing (Polars) with astronomical computations and FFT-based signal analysis. Designed for scalability and reproducibility, the system demonstrates how large-scale statistical pipelines can rigorously test hypotheses on noisy real-world data.
ML deployment · 2024
Lightweight pipeline for converting TensorFlow models into TensorFlow.js format, enabling browser-based inference and client-side deployment. The project focuses on model portability across runtime environments, demonstrating how trained models can be adapted for low-latency applications and privacy-preserving use cases without server-side inference.
Interactive demo
A collection of AI algorithms for classic games — minimax, and AlphaZero-style policy networks. Play against the AI directly in the browser for Tic-Tac-Toe, and Connect Four.
Nature Communications · 2022
Cell cycle inference in single-cell RNA-seq using RNA velocity and a circular autoencoder. Assigns a continuous transcriptional phase angle to each cell, revealing hidden oscillatory structure without external perturbations or cell sorting.
PNAS · 2019
Simulation of ribosome dynamics along mRNA transcripts using an inhomogeneous TASEP with codon-specific elongation rates. Combined with ribosome footprinting data, the model revealed that amino acid composition is as important a determinant of translation elongation speed as codon–tRNA adaptation — challenging a long-standing assumption in molecular biology.
Web application
A full-stack analytical application combining data processing, forecasting models, and interactive visualization. Designed to explore portfolio dynamics and simulate future scenarios.
Interactive demo
Interactive 3D visualisation of the night sky using real HYG star catalogue data and physically accurate spectral colours. Navigate a rendition of the observable universe in your browser with Three.js.
Background in computational biology with publications in Nature Methods, Nature Communications, PNAS and others.
Full list (12 papers, 1,020+ citations) → Google Scholar