Our clients operate in 14 countries. Their customers write support tickets in Portuguese, Thai, German, Vietnamese, and eight other languages — often mixed. Our current NLP system handles English well and falls apart everywhere else. We need an NLP engineer who has actually shipped multilingual classification and entity extraction at scale — not someone who knows the theory, but someone who has fixed models when they failed across language pairs.
Responsibilities
Build and fine-tune multilingual classification and extraction models
Design language detection and routing pipelines
Evaluate model performance per language and identify gaps
Collaborate with annotation teams on training data quality
Document model behaviour and limitations for client teams
Requirements
Strong experience with multilingual NLP (cross-lingual models, mBERT, XLM-R)
Python and Hugging Face Transformers
Experience with text classification, NER, or intent detection at scale
Familiarity with translation pipelines and language detection
Able to evaluate model performance across language pairs