NLP Engineer — News Intelligence

Yuki Yamamoto

Contract · Mid-level · London

This role is no longer accepting applications

Similar open roles you can apply to now:

About the role

We are a news intelligence company. We aggregate and process content from 85,000 news sources in 42 languages in real time. Our platform is used by communications teams, investors, government press offices, and corporate affairs functions who need to understand what is being written about topics, companies, and people — and why it matters. The NLP stack is the product. Topic classification, entity recognition, sentiment analysis, narrative clustering, and influence scoring — these are not features in a platform, they are the platform. We are hiring a contract NLP Engineer for a 9-month engagement to work on two specific improvements to our English-language pipeline: improving entity disambiguation in financial and political contexts where homonymous entities are common, and extending our narrative clustering model to handle emerging topics that were not represented in the training data. The work is applied and fast-moving. Our training data is real news, which means it is noisy, politically complex, and occasionally adversarial. We expect the engineer we hire to care about those properties, not just the benchmark numbers.

Responsibilities

Improve entity disambiguation for financial organisations and political figures in English-language news content
Extend narrative clustering to handle emerging topics through online learning or few-shot adaptation techniques
Build evaluation datasets and automated evaluation pipelines for both tasks
Work with the editorial and product teams to understand what disambiguation failures and clustering errors cost in terms of product quality
Document all model changes, data decisions, and evaluation methodology for handover to the permanent team

Requirements

4+ years of NLP engineering with production experience in entity recognition, classification, or text clustering
Hugging Face Transformers for fine-tuning on domain-specific text data — you understand when to fine-tune and when not to
Python at a professional standard — readable, maintainable code that the permanent team can extend after your contract ends
SpaCy for pipeline construction, rule-based extraction layers, and integration with transformer-based models
SQL for working with article metadata, entity resolution tables, and evaluation dataset management
Statistics for evaluation: precision-recall trade-offs in multi-class classification, inter-annotator agreement, and the discipline to build evaluation sets that measure what actually matters
Experience with multilingual NLP or low-resource language adaptation is an advantage

Benefits

$530 – $710 per day (inside or outside IR35 — to be confirmed at contract stage)
9-month initial contract
Fully remote with optional access to our London office
Real news data at a scale and variety that is unusual outside major media intelligence firms
Clear scope, defined deliverables, and a permanent team that is genuinely invested in the handover