Senior Data Scientist

Datumtg; Raleigh (NC); On-site; USD 120,000 - 170,000; Full time

Job summary

Datumtg in Raleigh, NC is seeking engineers to advance LLM-based information-extraction and NLP workflows. You will design prompts, build RAG and multi-step agents, and create robust data-prep pipelines with strong testing practices.

We value collaboration with SMEs to ensure factual accuracy, and you will balance quality, latency, and cost while maintaining clear, versioned code and documentation. A strong background in Python and ML is essential.

Qualifications

• Bachelor's degree or higher in CS/DS/Stats or related field.
• At least 3 years in data science, ML, or NLP with hands-on LLM apps.
• Strong Python, SQL, and data handling (JSON/XML).
• Experience with prompt design, structured outputs, and LLM evaluation.
• Familiar with Git, debugging, and automated tests.
• Ability to communicate findings clearly and manage risks.

Responsibilities

• Develop and improve LLM-based workflows for extraction, classification, summarization, and content generation.
• Design and evaluate prompts, RAG, and multi-step agent workflows.
• Build data-prep pipelines including parsing, transformation, validation, and caching.
• Create evaluation datasets and run controlled experiments and regression tests.
• Collaborate with SMEs to assess factual accuracy and translate feedback into improvements.
• Diagnose problems across data, model behavior, and application logic.
• Balance output quality, latency, and cost of approaches.
• Maintain tested, version-controlled code and documentation.

Skills

Python programming

SQL

LLM development

Git

Automated tests

Education

Bachelor's degree or higher in CS/DS/Stats

Tools

PyTorch

scikit-learn

AWS S3/EC2

JSON

XML

Linux

Job description

Location: Raleigh, NC || Onsite / Remote (EST Only)

Note: Preference will be given to candidates with Mandarin language proficiency.

Job Description:

Key Responsibilities

• Develop and improve LLM-based workflows for information extraction, classification, legal summarization, and content generation.
• Design and evaluate prompts, retrieval-augmented generation (RAG), and multi-step agent workflows.
• Build reliable data-preparation pipelines, including parsing, transformation, validation, caching, and recovery from interrupted processing.
• Create representative evaluation datasets and conduct controlled experiments, error analysis, and regression testing.
• Work with subject matter experts to assess factual accuracy, completeness, source support, and consistency, and translate feedback into generalizable improvements.
• Diagnose problems across data, retrieval, model behavior, and application logic.
• Balance output quality, latency, and cost when comparing technical approaches.
• Maintain tested, version-controlled code and clear documentation to support reproducibility and team handoffs.

Required Qualifications

• Bachelor’s degree or higher in computer science, Data Science, Statistics, Mathematics, or a related quantitative field, or equivalent practical experience.
• At least 3 years of relevant experience in data science, machine learning, or NLP, including hands-on experience developing LLM-based applications.
• Strong Python programming skills and experience with SQL and structured or semi-structured data, such as JSON and XML.
• Practical experience with prompt design, structured outputs, retrieval, and systematic LLM evaluation.
• Understanding of statistical evaluation, train/validation/test separation, data leakage, and appropriate performance metrics.
• Experience using Git, debugging existing code, and writing automated tests.
• Ability to communicate findings clearly, incorporate feedback, and proactively raise delivery risks.

Preferred Qualifications

• Experience with traditional machine learning, including feature engineering, classification, tree-based models, and imbalanced datasets.
• Familiarity with scikit-learn, PyTorch, or comparable frameworks.
• Experience fine-tuning transformer models for classification, information extraction, or sequence labeling.
• Familiarity with AWS, particularly S3 and EC2, and Linux or GPU-based training environments.
• Experience with legal documents, editorial workflows, or other accuracy-sensitive domains.

What are the top 3 must-have skills?

• Applied LLM development: demonstrated experience building and evaluating workflows involving prompts, retrieval, and structured generation.
• Python and data engineering: ability to independently build, debug, and maintain reliable data-processing and model-integration code.
• Experimental design and evaluation: ability to compare baselines, analyze errors, validate improvements, and communicate actionable findings.