Staff Data Engineer, Multilingual NLP & Open Models

Socket.dev; San Francisco (CA); On-site; USD 130,000 - 170,000; Full time

Benefits offered by this job

Compensation

Stock options

Healthcare

Meals provided

Parental leave

Unlimited vacation

Visa sponsorship

Team events

Job summary

Reflection is a research lab making intelligence open and accessible for everyone to use, customize, and build on. We design and operate large-scale multilingual data pipelines, source, clean, deduplicate, identify language, and normalize scripts across high- and low-resource languages, with a focus on quality and scalability.

As part of a small, fast-moving team you will lead small research projects, build evaluation sets, and collaborate with pre-training, mid-training, and post-training

Qualifications

• Strong software engineering fundamentals and comfort processing web-scale datasets in distributed environments.
• Experience building large-scale data pipelines for language models, machine translation, speech, or search — ideally covering more than one language.
• Fluency or working proficiency in at least one language other than English, and genuine curiosity about how languages differ.

Responsibilities

• Design and operate large-scale multilingual data pipelines — sourcing, cleaning, deduplication, language identification, and script normalization across high- and low resource languages.
• Define and enforce quality bars for multilingual corpora, including translation quality, cultural fidelity, toxicity, and contamination checks.
• Design and run scientific experiments to advance our understanding of scaling large language models to improve multilingual data efficiency.
• Lead small research projects independently while collaborating on larger initiatives.
• Build evaluation sets and diagnostics that expose where model behavior degrades by language, register, or domain, and close those gaps with targeted data.
• Work with pre-training, mid-training, and post-training teams to land measurable, step-function improvements in multilingual capability.

Skills

Software engineering

Data pipelines

Multilingual experience

Measurement-driven

Trade-offs

Ambiguity tolerance

Education

MS/PhD in CS/ML

Tools

Tokenizer design

Job description

Reflection is a research lab making intelligence open and accessible for everyone to use, customize, and build on. We design and operate large-scale multilingual data pipelines, source, clean, deduplicate, identify language, and normalize scripts across high- and low-resource languages, with a focus on quality and scalability.

As part of a small, fast-moving team you will lead small research projects, build evaluation sets, and collaborate with pre-training, mid-training, and post-training