Pre-Training Data ML Research Engineer — Equity & Impact

Liquid AI; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Benefits offered by this job

Unlimited PTO and Refill Days

Equity opportunities

Comprehensive health coverage

Paid leave for family and medical

Job summary

Liquid AI, spun out of MIT CSAIL, is seeking a data-centric ML engineer in San Francisco to own end-to-end pre-training data pipelines. You will source, filter, and synthesize data for foundation model training, building pipelines that scale with demands of large models.

You will run experiments to quantify data changes, reason about data interactions with model size and architecture, and ensure reproducible training runs. We value hands-on data work and a track record of shipping data solutions.

Qualifications

• Deep, hands-on pre-training data work you did yourself.
• Ability to reason from a data decision to its effect on training.
• Experiments you designed to test data changes with baselines, scaling ladders, and held-out loss.

Responsibilities

• Source web-scale and public datasets and build pipelines to filter, deduplicate, and decontaminate them.
• Design data mixtures and curricula for pre-training and midtraining and decide what goes into each run.
• Build synthetic data generation systems using LLMs, structured prompting, and distillation.
• Run proxy experiments and ablations to measure data changes and how they affect model quality.
• Build evaluation sets and maintain data lineage for train/test reproducibility.

Skills

Pre-training data work

Data impact reasoning

Experiment design

Production data pipelines

Python

PyTorch

Education

MS in ML/CS

PhD in ML/CS

BS with 5+ years experience

Tools

PyTorch

Python

Job description

Liquid AI, spun out of MIT CSAIL, is seeking a data-centric ML engineer in San Francisco to own end-to-end pre-training data pipelines. You will source, filter, and synthesize data for foundation model training, building pipelines that scale with demands of large models.

You will run experiments to quantify data changes, reason about data interactions with model size and architecture, and ensure reproducible training runs. We value hands-on data work and a track record of shipping data solutions.