Applied AI systems
Machine learning, RAG, retrieval and LLM evaluation designed around explicit quality criteria and operational constraints.
- Applied machine learning
- Retrieval and evaluation
- NLP and document workflows
Independent Data Scientist & ML Engineer
I design and evaluate applied ML, RAG and data systems.
My work covers RAG evaluation, data engineering, data quality and analytics. I make the criteria, evidence and limitations explicit so the work can be tested and reviewed.
Working across
- Applied machine learning
- RAG Reliability Audit
- Data engineering and quality
- Analytics and BI
01 / Capabilities
What I work on
Practical areas I take on, organised around delivery problems rather than disconnected tools or tutorial exercises.
Data foundations
Pipelines, data quality and analytical workflows built for reproducibility, observability and clear ownership.
- Data engineering
- PySpark and Databricks
- Data quality and validation
Decision analytics
Analytics, forecasting and BI work that turns technical outputs into information people can evaluate and act on.
- Analytics and BI
- Forecasting
- Technical communication
02 / Approach
A delivery process that makes assumptions and trade-offs visible.
- 01
Frame
Clarify the business problem, users, constraints, evidence and definition of success.
- 02
Build
Implement the smallest coherent solution with typed boundaries, reproducible tooling and maintainable structure.
- 03
Evaluate
Test behaviour, data quality and model or retrieval performance using criteria suited to the actual risk.
- 04
Communicate
Explain results, limitations, operating requirements and next decisions to technical and non-technical stakeholders.
03 / Evidence
What the published work shows.
Published case studies document the problem, approach, evaluation method, trade-offs and production considerations, including what the evidence does and does not support.
RAG Reliability
Synthetic portfolio study
A proprietary product direction for deciding when a RAG system should answer or defer. The evidence published here comes from a synthetic answer/defer study.
Data quality audit
Interactive demo in development
Profiling, validation rules, anomalies and decision-ready reporting.
Pipeline modernisation
Verified professional work
Reproducible data workflows, testing, observability and handover.
Documents
CVs, academic work and professional profile
Review the work, then tell me about the problem.
Look through the services and published case studies, then get in touch to discuss the current system, the decision that matters and the constraints involved.
For small-business projects, I am also building MigaDigital, a separate initiative focused on practical data, automation and AI work. Visit MigaDigital (Website currently in Spanish)