Senior Data Engineer / Machine Learning Engineer (Product Intelligence & Matching)

You will design and build a system that can:

• Interpret unstructured product descriptions
• Extract key attributes (brand, model, size, etc.)
• Categorise products into structured taxonomies
• Identify duplicate or equivalent products across datasets

This is a technically challenging role involving text processing, data engineering, and intelligent matching, with real-world impact across enterprise and government datasets., * Design scalable pipelines for processing large product datasets

• Build robust text parsing and normalisation systems
• Develop classification models for product categorisation
• Implement advanced matching algorithms combining:
• Rules-based logic
• Statistical similarity
• Machine learning approaches
• Work closely with product and leadership teams to shape this new capability
• Ensure solutions are fully deployable on-premise (no cloud dependency)

• Strong experience in Python or C# (.NET)
• Proven experience in data matching, entity resolution, or record linkage
• Experience working with text-heavy datasets
• Solid understanding of:
• NLP techniques
• Similarity scoring and ranking
• Data modelling and transformation
• Ability to design solutions from first principles

Desirable Skills:

• Experience with product catalogue data or e-commerce datasets
• Experience with classification and taxonomy systems
• Familiarity with local/offline ML models
• Knowledge of data quality tools or MDM systems

• Examples of similar work (entity matching, NLP, product data, etc.)
• A brief outline of how you would approach this problem
• Your preferred tech stack for implementation

We are looking for someone who can think, design, and deliver - not just code to a spec.

Job Types: Freelance, Temp to perm

Pay: £25.00-£35.00 per hour

Application question(s):

• How would you approach matching two product descriptions that are written very differently but refer to the same item?
• What techniques would you use to extract structured attributes (e.g. brand, size, colour) from messy product descriptions?
• How would you design a matching score between two products?
• How would you avoid comparing every product to every other product (performance problem)?