Research Engineer Pretraining Data

14 hours agoSalary: 150K - 350KSan Francisco, USA Remote Full Time Ai Jobs by Prime Intellect

Skills

Data Contamination Data Processing Data Provenance Dataset Dataset Versioning Deduplication Distributed System Document Extraction Midtraining Pretraining Python Ray Spark Synthetic Data Text Classification Tokenization Web Crawling

Candidate Availability

Remote · San Francisco · Required

Required and preferred rules are kept separate and reflect the wording in the original posting.

About the Role

You will source and curate training data, build scalable processing pipelines, assess data quality and contamination, develop sampling strategies, prepare long-context, reasoning, and coding data, run controlled experiments, and maintain reproducible datasets with clear provenance and versioning.

Requirements

Python programming skills
Experience building reliable data-processing or distributed systems
Experience with large datasets
Understanding of storage, compute, throughput, and reproducibility tradeoffs
Ability to diagnose data-quality issues
Ability to design measurable checks and experiments
Experience owning technical problems end to end

Responsibilities

Find, evaluate, and collect training data from web, code, and technical sources
Build pipelines for extraction, cleaning, deduplication, and filtering
Develop data-quality, coverage, and contamination checks
Work on pretraining and midtraining data mixtures and sampling strategies
Prepare and evaluate data for long-context learning, reasoning, and coding
Run controlled data experiments and evaluate model outcomes
Make large-scale processing efficient, reproducible, and recoverable
Maintain source provenance and dataset versioning

Benefits