Machine Learning Engineering
Building analytical pipelines, scoring workflows and model-backed systems for noisy, changing or incomplete data environments.
Profile
I build machine-learning and analytical systems for noisy, incomplete fast-moving data.
My work focuses on market intelligence, anomaly detection and probabilistic decision
system reliability. The common theme is turning unstable signals into reasons that can
be inspected before it is trusted.
My independent portfolio is organised around 3 core areas: market intelligence, reliability monitoring and
real world infrastructure.
The wider direction is evidence-led audits across crypto, property and AI infrastructure: testing
what supports a claim, what weakens it, what is missing and what should be checked next.
core focus
Building analytical pipelines, scoring workflows and model-backed systems for noisy, changing or incomplete data environments.
Testing uncertainty, instability, anomalies and failure points before model outputs are used in operational or financial decisions.
Applying probabilistic modelling, signal analysis and diagnostic evidence chains to markets, risk and decision systems.
Technical ownership
Built a reproducible cross-domain analytical framework across 23,951 grid cells on a 25 km national raster, integrating manufacturing, infrastructure, port-accessibility and satellite-derived features.
Methods: PCA, UMAP, HDBSCAN, ARI, local concordance, bootstrap stability, Moran's I and LISA.
View research →Tested whether learned structures were stable, locally consistent and decision-useful before treating them as evidence.
Measured output: 10 repeated UMAP seeds, 25 bootstrap runs per method, 300 pairwise ARI comparisons, 999 Moran permutations and 499 LISA permutations.
View projects →Converted model disagreement into outputs that can be inspected through maps, metrics, filters, dashboards, tooltips and cell-level evidence trails.
Measured output: UMAP produced 9 clusters with 93.6% mean local stability; PCA produced 10 clusters with 38.8% noise; satellite clustering produced 3 clusters with 65.5% noise.
Experience
A compact view of the work behind my independent portfolio, supported by property, commercial and financial-services experience.
Skills