Staff Data Engineer, Pretraining Data & Experiments

Speedrun Talent Network; San Francisco (CA); In loco; USD 300.000 - 400.000; Tempo pieno

Vantaggi offerti da questo lavoro

Equity grant

Competitive benefits

Descrizione del lavoro

Mirendil is a tech-first company advancing frontier AI R&D by building in-house models and running end-to-end experiments. You will set training data directions, run ablations, and supervise data pipelines that scale to trillions of tokens.

You’ll work with researchers on architecture and training runs, ensuring data quality, provenance, and reproducibility as the corpus grows. A base salary and meaningful equity are offered, with competitive benefits.

Competenze

• Shipped a data change in a large pretraining run.
• Experience building data pipelines at scale (trillions of tokens).
• Proficient Python and ML framework tooling (PyTorch or JAX).

Mansioni

• Build data pipelines for pretraining and manage code and docs.
• Improve data extraction, filtering, and deduplication of public data.
• Design ablations and data scaling experiments for model training.
• Coordinate with researchers to run experiments and evaluate results.
• Extend data mixtures to new scientific domains.
• Maintain data provenance and reproducibility as corpus grows.

Conoscenze

Data pipelines

Python

Spark

Ray/Beam

PyTorch/JAX

Strumenti

Spark

Ray

Beam

Descrizione del lavoro

Mirendil is a tech-first company advancing frontier AI R&D by building in-house models and running end-to-end experiments. You will set training data directions, run ablations, and supervise data pipelines that scale to trillions of tokens.

You’ll work with researchers on architecture and training runs, ensuring data quality, provenance, and reproducibility as the corpus grows. A base salary and meaningful equity are offered, with competitive benefits.