Lead ML Engineer (Document AI NLP) at Intelance, London, £500-£800 per day

Intelance is a specialist architecture and AI consultancy working with clients in regulated, high-trust environments (healthcare, pharma, life sciences, financial services). We are building a lean senior team to deliver an AI-assisted clinical tool for a UK-based organisation in human genetic testing. We are looking for a Lead ML Engineer who can turn messy real-world documents into reliable, explainable model outputs. This is a contract / freelance role, part-time (2-3 days/week), working closely with our AI Solution Architect and Data Engineer.

•  
• Design and implement the ML/NLP core of an AI-assisted marking tool that:

○ Ingests clinical-style reports (PDF/Word) via an OCR + parsing pipeline

○ Extracts relevant content and features

○ Applies a hybrid scoring approach (rules + LLM / transformer models)

○ Outputs scores, rationales, and confidence levels.

•  
• Build and iterate prompting / few-shot setups and rule layers so that model behaviour is consistent, predictable, and easy to explain to assessors.
• Work with the Data Engineer to define and consume clean structured inputs from the OCR/pipeline (schemas, validation checks, logging).
• Implement evaluation pipelines: ground-truth comparisons, error analysis, per-criterion metrics, and drift/robustness checks.
• Optimise models for accuracy, stability, and cost (latency, token usage, throughput) within agreed constraints.
• Support the architect and compliance lead in designing explainability and audit: what is logged, what is shown to assessors, and what evidence is retained for validation.
• Package models behind clean interfaces (e.g. Python services, APIs, batch jobs) so they can be integrated with the rest of the system.
• Participate in technical workshops with the client to walk through behaviour on real examples and collect feedback.
• Document your work clearly: experiments, model choices, prompt patterns, known limitations, and recommended operating boundaries.

Must-have

•  
• 4+ years of hands-on Machine Learning / NLP engineering experience (not just research).
• Strong Python skills and experience with at least one modern ML/NLP stack (PyTorch, TensorFlow, HuggingFace, spaCy, etc.).
• Practical experience with document AI / text processing: PDFs, OCR outputs, long-form text, classification or scoring of documents.
• Solid understanding of LLMs and prompt-based workflows (e.g. OpenAI/Azure OpenAI, Anthropic, or similar) and how to mix them with rules / traditional models.
• Experience building evaluation pipelines: test sets, metrics, error analysis, and data-driven model selection.
• Comfort working in environments where explainability, auditability, and consistency matter more than bleeding-edge novelty.
• Ability to work independently in a small senior team, take ownership of a problem, and communicate clearly about trade-offs.
• Available for 2-3 days per week on a contract basis, working largely remotely in UK or close European time zones.

Nice-to-have

•  
• Prior work in healthcare, life sciences, clinical reporting, or regulated industries.
• Experience with Azure (Azure ML, Azure Functions, Azure OpenAI, blob storage) or other major cloud providers.
• Exposure to validation or quality frameworks (e.g. GxP, ISO 15189, UKAS, NHS IG).
• Familiarity with MLOps practices (versioning, deployment, monitoring), even at a lightweight level.

•  
• Real impact: build a production AI system that will support external quality assessment in human genetic testing.
• Lean, senior team: work directly with an AI Solution Architect, experienced Data Engineer, and the leadership team – quick decisions, minimal bureaucracy.
• Remote-first, flexible: work from anywhere compatible with UK business hours, with a planned load of 2-3 days/week.
• Contract / freelance: competitive day rate, with the potential to extend into further phases and additional schemes if the pilot is successful.
• Opportunity to help define reusable ML/NLP components that Intelance will deploy across multiple regulated AI projects.

We review every application personally. If there’s a good match, we’ll set up a short call to walk through the project, expectations, and next steps.