Applied Scientist/Research Engineer, LLM Training Data

Description

Propio Language Services is a top-five global language services provider and the fastest-growing company in the industry. Operating at nine-figure scale across healthcare, legal, and other sectors, Propio delivers high-quality, real-time multilingual interpretation, translation, and localization services. Driven by cutting-edge technology and exceptional service, we create seamless experiences that bridge communication gaps across languages, cultures, and communication channels.

Propio is hiring an Applied Scientist/Research Engineer, LLM Training Data to own the data strategy, curation pipelines, annotation workflows, and evaluation datasets that power our multilingual AI systems. This is a hands-on technical role for someone who understands how to manage the full AI data lifecycle, from acquisition, curation, annotation, and quality control to evaluation datasets and post-training data, to directly improve LLM performance. The ideal candidate can build scalable data pipelines, design high-quality annotation and QA processes, identify model failure modes, and close performance gaps through targeted data acquisition, curation, and synthetic data generation.

Key Responsibilities:

• Define the end-to-end data roadmap for multilingual and multimodal AI systems, including text, speech, translation, interpretation, low-resource languages, and agentic AI workflows.
• Design and build dataset curation pipelines for training, post-training, and evaluation, including cleaning, deduplication, filtering, PII redaction, quality scoring, sampling, balancing, and versioning.
• Create annotation schemas, labeling guidelines, QA rubrics, golden datasets, and reviewer workflows for multilingual, speech, translation, and vision data.
• Build evaluation datasets and benchmarks, analyze model failure modes, and translate performance gaps into targeted data improvements.
• Support post-training data workflows such as SFT, instruction tuning, preference data, RLHF/DPO-style data, reward model data, and synthetic data generation.
• Use modern annotation tools and AWS-based data infrastructure to scale secure, traceable, and compliant AI data workflows.

Requirements

Qualifications:

• Master’s degree in Computer Science, Machine Learning, Data Science, Computational Linguistics, Linguistics, Statistics, or a related field, or equivalent practical experience. 
• 4+ years of experience in AI data, ML data operations, NLP data engineering, applied ML, speech/translation data, or LLM data workflows. 
• Strong hands-on experience with Python, SQL, and dataset curation pipelines. 
• Experience with annotation workflows, QA rubrics, evaluation datasets, or human-in-the-loop data processes. 
• Familiarity with multilingual NLP, speech data, translation data, low-resource languages, conversational AI, or agentic AI datasets. 
• Working knowledge of AWS data and ML tools such as S3, Glue, SageMaker, Bedrock, Lambda, Step Functions, EKS/ECS, IAM, or KMS. 
• Strong communication skills and ability to work with ML engineers, applied scientists, product teams, linguists, data teams, and vendors. 

Preferred Qualifications:

• PhD in Computer Science, Machine Learning, NLP, Computational Linguistics, Data Science, Statistics, or a related field. 
• Experience with LLM post-training workflows such as SFT, instruction tuning, preference data, RLHF, DPO, reward modeling, or evaluation data generation. 
• Experience with synthetic data generation, active learning, weak supervision, LLM-as-judge workflows, or automated data quality scoring. 
• Experience with modern annotation and data platforms such as Labelbox, Scale AI, Prodigy, Argilla, Snorkel, Humanloop, or custom internal tooling.

#LI-JS1