Skip to content

Services · Capabilities index

End-to-end ML & AI, from architecture to production.

I’m a senior data scientist and AI engineer with a PhD and deep industry experience. I learned end-to-end delivery by doing all of it, from model architecture and MLOps to agentic AI, RAG, and stakeholder management. Some of what I built is still in use today.

Machine Learning & Forecasting

01

Machine Learning Solutions

Custom models for business-critical decisions, applying classification, regression and clustering to use-cases such as customer churn and retention, risk scoring, value estimation and customer segmentation. I have delivered a customer-retention model with a statistically significant, value-quantified uplift.

XGBoostCatBoostScikit-learnPyTorchPyMC
02

Forecasting & Demand Planning

Time series forecasting for planning and operations, covering demand, capacity and financial KPIs across industrial and subscription/SaaS businesses. I built financial KPI forecasting and raw-material demand forecasting, and set the time series practices a global team still relies on.

ProphetPyMCDARTSNixtlaStatsmodels
03

Predictive Maintenance & Anomaly Detection

Condition monitoring for industrial equipment, applying classification, causal models, anomaly detection and clustering to failure detection, root-cause analysis, predictive maintenance and event grouping. I have delivered equipment-monitoring models in production, with drift detection.

Scikit-learnXGBoostStumpyRupturesEvidently
04

Recommender Systems & Personalization

Recommendation and next-best-action systems for product, content and operations. I have built recommender solutions, including a product design recommender, and apply ranking and personalisation across customer and operational use cases.

PythonScikit-learnEmbeddingsRanking

Causal & Decision Science

01

Causal Inference & Experimentation

Understanding what actually drives an outcome, not just what correlates with it: A/B testing, uplift modelling and causal analysis. I built the A/B testing framework behind a production retention model and use Bayesian and causal methods throughout.

PyMCCausalNexDoWhyA/B Testing
02

Marketing Analytics & MMM

Technical lead on a flagship Media Mix Modeling programme that captured tens of millions in value in its first year and is still used for strategic decisions. Bayesian attribution, ROI measurement and budget optimisation.

PyMCBayesian ModelingCausal InferenceMarketing Analytics
03

Logistics & Operations Optimization

Optimisation for supply chain and operations: routing, allocation, scheduling and scenario simulation. I have delivered truck-utilisation and route optimisation with delivery-time estimation, and advised on the approach, scoping and phasing of a transportation optimiser prototype.

PythonNetworkXPolarsScikit-learnPyMC

GenAI, Agents & LLM Applications

01

Agentic AI & RAG Applications

Autonomous agents and retrieval-augmented systems built for production. I delivered a RAG knowledge-retrieval system, deployed to production, and a management reporting agent with execution tracing to debug how it retrieves and reasons.

LangGraphPydanticAILangChainChromaDBFAISS
02

GenAI Workflow Automation

Automating and streamlining business processes with LLMs and agents: document-heavy workflows, reporting, triage and internal copilots. I design these as small, verifiable steps so they behave reliably once in production.

LangGraphPydanticAIMLflow Tracing
03

Document Intelligence & Extraction

Turning unstructured documents into structured, reliable data. I extract fields, tables and entities from contracts, invoices, forms and reports, with classification and validation so the output is ready for downstream systems. This is the extraction layer beneath document-heavy workflows, separate from question-answering.

PythonLLM ExtractionPydanticValidation
04

AI Reliability & Evaluation

Making LLM systems trustworthy in production: execution tracing, LLM-as-a-judge evaluation, regression testing and observability. I trace how an agent retrieves and reasons, then test its answers against baselines.

MLflow TracingLLM-as-a-judgeCustom Evals
05

Private & Self-Hosted LLM Deployment

Running open-weight models on your own infrastructure so sensitive data never leaves the organisation. A practical option for regulated and privacy-sensitive work where public APIs are not acceptable.

LM StudioLlamaOpen-weight modelsOn-prem

Advisory, Strategy & Standards

01

Fractional ML/AI Lead

Senior technical leadership without a full-time hire. I set architecture, guide delivery and make the hard technical calls. I have led a 15+ person interdisciplinary team and owned high-impact programmes end to end.

ArchitectureDelivery leadershipTechnical strategy
02

AI Strategy & Use-Case Discovery

Finding where AI and ML actually create value: opportunity mapping, feasibility, prioritisation and scoping before a line of code is written, grounded in what survives contact with production.

Use-case discoveryFeasibilityRoadmapping
03

Architecture & Technical Due Diligence

An independent review of AI and ML systems and plans: architecture, build versus buy, vendor and approach sanity checks, and risk assessment before you commit budget.

Architecture reviewDue diligenceRisk assessment
04

Data Science Maturity Assessment & Standards

How good is your data science, really, and how do projects compare? I define the criteria and methodology to evaluate product maturity and benchmark projects against each other, then turn the findings into concrete standards and good practices.

Maturity assessmentBenchmarkingStandards & best practices

Data, MLOps & Enablement

01

Data Engineering & Big Data

Large-scale processing and distributed computing: PySpark on Databricks, cloud data lakes and warehouses, query engines, and ETL pipelines with orchestration across AWS and Azure.

PySparkDatabricksAWSAzureGitHub Actions
02

MLOps Implementation

End-to-end pipelines for reliable deployment and maintenance. I have built deployment blueprints and established MLOps practices in production: CI/CD, drift detection, data quality, monitoring and reproducibility.

DockerKubernetesMLFlowGitHub ActionsAzure DevOpsDatabricks
03

Entity Deduplication & Record Linkage

Cleaning duplicate records that accumulate across systems: fuzzy matching, phonetic algorithms, probabilistic ML record matching and active learning, for customer master data, product catalogs and spare-parts master data, at the scale of millions of records. I applied it to spare-parts master data to improve pooling and reduce global stock.

PandasPolarsPySparkRecordLinkageDedupeActive learning
04

Python Training & Mentoring

Workshops and mentoring on Python, data science and ML engineering: software engineering best practices, testing and code quality, modern tooling, and team enablement. Guided by the Zen of Python.

Best practicesTestingTeam enablement

Let’s work together.

Build from scratch, lead a complex delivery, or connect the technical and business sides. I bring that range of experience.

Get in touch