Chris M.

I turn tangled data and hard modelling problems into things you can actually use: a production-grade model, a deployed app, a clear answer to a complex question. I hold a PhD in Complexity Sciences and a first in Theoretical Physics, and I've spent ~15 years as a researcher and consultant building this stuff for real. Most of that work has been across fields that don't usually talk to each other: critical-care medicine, theoretical ecology, economics, education, and industry. This extensive interdisciplinary experience has let me use the same methods time and again (machine learning, agent-based simulation, network analysis, optimisation) in different application areas. As such, I can recognise which one your problem actually requires, and can rapidly learn and deploy new technical methodologies as needed. A few concrete things: I built and deployed a Flask decision-support app for ICU discharge that came out of a machine-learning study (published in BMJ Open). I've written agent-based models of illegal fishing and GPU-accelerated simulations with large speed-ups over the naive version. I've done reinforcement-learning and optimisation for allocating people at organisational scale, control software for off-grid sanitation hardware, and Streamlit data tools for public-good projects. My research has appeared as first-author work in Nature Communications, PNAS, and BMJ Open. Two things I'm reliably good at: picking up an unfamiliar technique and getting it working quickly, and explaining a complicated result to the people who have to act on it. I work in full-stack Python and am comfortable across the whole pipeline: extraction and cleaning, exploratory analysis, model building, and deployment. I run a small consultancy, Rusty Data, focused on helping organisations get value out of their data. If you've got a modelling problem, a pile of data you suspect is useful, or an idea you want pressure-tested, send me a message with a bit of detail and I'll tell you how I'd approach it. How I can help: Build and deploy ML models - classifiers, deep learning (CNN/RNN), predictive tools - end to end, not just a notebook. LLM and agentic work: RAG systems, NLP pipelines, and automation of the tedious analysis you'd rather not do by hand. Simulation and modelling: agent-based models, dynamical systems, and network analysis, including GPU-accelerated versions when speed matters. Optimisation and decision support: numerical optimisation, reinforcement learning, and bespoke apps that turn a model into something a team can use. Data wrangling and exploratory analysis on messy, large, or HPC-scale datasets. Data visualisation and interactive dashboards (d3.js, Tableau, Streamlit, Bokeh) that make a result legible. A second opinion - feasibility reviews, method selection, and clear write-ups for technical or non-technical stakeholders. Skills & tools: Languages: Python (full-stack), SQL (MySQL/Postgres), C++, R, MATLAB/Octave, NetLogo, Unix/shell Python stack: TensorFlow, Keras, PyTorch, scikit-learn, Pandas, SciPy, NumPy, Matplotlib, Bokeh, Streamlit, Flask, Django, Jupyter Infrastructure: big data, HPC, GPU acceleration, cloud compute Visualisation: d3.js, Tableau, Streamlit, Bokeh, Matplotlib Methods: ML (classifiers, CNN/RNN deep learning, deployment) · LLMs (NLP, RAG, agentic workflows, automation) · reinforcement learning · agent-based & dynamical-systems modelling · network analysis & inference · dimensionality reduction (PCA, t-SNE, kernel) · structural equation models & causal inference · numerical optimisation (simulated annealing, basin hopping) · full-stack software development · data visualisation