Multilingual training data specialist

About the role

You will own the linguistic quality of training datasets built for machine learning clients expanding across Latin America. This includes designing label schemas that capture pragmatic variation between Spanish variants, running cross-variant calibration across annotation desks, and working directly inside client training pipelines so annotation priorities respond to model performance. This role sits at the intersection of linguistics and machine learning. You do not need to build models. You do need to understand why a model trained on Castilian Spanish fails in the Caribbean, and to design the annotation approach that fixes it.

Key responsibilities

• Design and maintain annotation label schemas, including pragmatic dimensions such as directness, politeness register, urgency signalling and sentiment intensity
• Document divergence taxonomies describing how usage differs systematically between Spanish variants
• Run weekly cross-variant calibration sessions across annotation desks to maintain dataset consistency
• Define and monitor inter-annotator agreement thresholds per variant, not pooled
• Work inside client machine learning pipelines, setting annotation priorities weekly against model uncertainty and measured performance
• Implement active learning approaches that concentrate annotation effort where the model is weakest
• Analyse model performance by language variant and identify the linguistic cause of performance gaps
• Design and run quality audits on delivered datasets
• Train and develop linguistic leads and annotation specialists
• Author client-facing documentation and present findings to client machine learning teams
• Contribute to solution design and estimation for new annotation engagements
• Maintain data handling compliance under Law 172-13 and client-specific protocols
• Stay current with developments in multilingual NLP evaluation and annotation methodology

Required qualifications

• Native Spanish with demonstrable knowledge of Latin American regional variation
• English at CEFR C1, assessed before offer. This role is client-facing and conducted substantially in English.
• Three or more years in linguistics, computational linguistics, data annotation, NLP quality or a closely related field
• Demonstrable understanding of annotation methodology: schema design, inter-annotator agreement, calibration, quality measurement
• Working familiarity with machine learning evaluation concepts: precision, recall, F1, confusion analysis
• Comfort with data tooling: spreadsheets at an advanced level, and ideally SQL or Python for dataset analysis
• University degree in linguistics, computational linguistics, languages, computer science or a related field
• Availability to work onsite in Punta Cana
• Valid Dominican ID or work permit

Preferred qualifications

• Postgraduate study in linguistics, computational linguistics or NLP
• Python for data analysis: pandas, numpy, basic scripting
• Prior experience with annotation platforms: Label Studio, Prodigy, Scale, Labelbox or equivalent
• Experience with RLHF, model evaluation or red-teaming
• Additional languages: Haitian Creole, Portuguese, French
• Prior experience working with distributed annotation teams
• Published work or academic research in dialectology, sociolinguistics or pragmatics

About Corpshore Dominicana

Corpshore Dominicana is the Dominican delivery arm of Corpshore Solutions Corporation, Toronto. We deliver BPO, IT and AI services from four Dominican cities, and we promote from within.

Apply for this role