Research Engineer - Pretraining Data

Prime Intellect is building the open superintelligence stack: the infrastructure frontier AI labs build internally, made available to every ambitious AI team.

Our platform, Lab, unifies compute, environments, evaluations, secure sandboxes, high-performance training, and deployment into one full-stack system for post-training at frontier scale - from SFT and RL to tool use, agent workflows, and continuously improving production models. We are building open frontier AI: open-source models trained end to end for long-horizon tasks like autonomous research, and the full-stack platform our own research team uses to build them. The next generation of AI companies, enterprises, and research teams do not just need more GPUs. They need the ability to turn their own workflows, tools, data, and feedback loops into superintelligence they own.

Prime Intellect has raised $150M in total funding from Founders Fund, Radical Ventures, NVIDIA, and exceptional AI, infrastructure, and enterprise operators, including Andrej Karpathy, Dwarkesh Patel, and leaders and founders from Ramp, Perplexity, Harvey, Mercor, Zapier, Datadog, Cognition, OpenAI, Thinking Machines, Together AI, SemiAnalysis, LangChain, Browserbase, Cloudflare, Sierra, Databricks, Airbnb, OpenRouter, Standard Intelligence, Fleet, Core Auto, and more. We are looking for people who want to build at the intersection of frontier research, real infrastructure, and go-to-market for a category that does not fully exist yet.

Build the data behind Prime Intellect’s next generation of open models. You’ll find valuable sources, turn them into training-ready corpora, and work with researchers to test what actually improves the model. Your work will shape both broad pretraining and the data used to develop reasoning, coding, and long-context capabilities.

• →

  Find, evaluate, and collect valuable training data from web, code, and technical sources.

• →

  Build scalable pipelines that turn raw data into training-ready corpora through extraction, cleaning, deduplication, and filtering.

• →

  Develop checks for data quality, coverage, and contamination, and use them to improve collection and processing.

• →

  Work with researchers on data mixtures and sampling strategies for pretraining and midtraining.

• →

  Prepare and evaluate data for long-context learning, reasoning, and coding, testing synthetic data where it proves useful.

• →

  Run controlled data experiments and use model evaluations to decide which sources, mixtures, and processing choices are worth scaling.

• →

  Make large-scale data processing efficient, reproducible, and recoverable, with clear source provenance and dataset versioning.

• Strong programming skills in Python and a record of building reliable data-processing or distributed systems.

• Experience working with large datasets and managing the practical tradeoffs of storage, compute, throughput, and reproducibility.

• Ability to diagnose data-quality issues and translate research goals into measurable checks and experiments.

• Comfort owning an evolving technical problem end to end and working closely with researchers to test assumptions and improve the system.

Relevant experience may include:

• Web crawling, document extraction, code-data collection, or large-scale corpus curation.

• Deduplication, text classification, quality scoring, synthetic data, or training-data contamination analysis.

• Data-mixture experiments, curriculum learning, or measuring how dataset changes affect model performance.

• Experience preparing datasets for language-model pretraining, continued pretraining, or midtraining.

• Work with distributed processing frameworks such as Spark, Ray, or comparable systems.

• Experience with multilingual, mathematical, scientific, or repository-level code data.

• Familiarity with tokenization, training data loaders, and the needs of large model-training runs.

Depth in data acquisition, scalable processing, or data research can be a strong starting point. We don’t expect every candidate to have worked across all of them.