About The Position
Liquid AI is hiring a research-minded ML engineer for its Pre-Training Data team. The team determines what Liquid Foundation Models learn by sourcing, filtering, synthesizing, and mixing data, as well as deciding the order in which they train on it. The role treats data as a research problem: you will work with pre-training researchers to test data hypotheses and build the pipelines that bring validated decisions into training runs. You will own ambiguous problems end to end, from forming a hypothesis and designing an experiment to implementing production pipelines and measuring the effect of data changes on model quality. You will also maintain trustworthy evaluation sets and traceable, reproducible data lineage. The role is suited to someone who combines deep hands-on pre-training data expertise with strong engineering execution and sound reasoning about how data interacts with model scale, architecture, and training dynamics.
Requirements
- Deep, hands-on experience personally performing pre-training data work, including web-scale sourcing and curation, synthetic data generation, mixing and curriculum design, deduplication, and decontamination.
- Ability to reason about how data decisions affect training, including interactions with model size, architecture, and training dynamics, why more data can hurt, and which changes matter as models scale.
- Experience designing experiments to test data changes, such as baselines, scaling ladders, small-scale proxies, and held-out loss, and explaining what a change did and why.
- Experience building and owning production data pipelines, not only research prototypes.
- Solid ML fundamentals, including training, evaluating, and iterating on models in Python; PyTorch is preferred.
- At least 3 years of relevant experience with an M.S., 1 year with a Ph.D., or 5 years with a B.S.
Nice To Haves
- Experience with distillation, long-context, or architecture-specific data strategy.
- Experience designing evaluation frameworks and benchmarks for pretrained models.
- Data work across both pre-training and post-training.
- Experience with small, efficient models or on-device deployment.
- Startup experience or a track record of connecting model work to product needs.
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 data goes into each run.
- Build synthetic data generation systems using LLMs, structured prompting, and distillation.
- Run proxy experiments and ablations to measure the effects of data changes on model quality and decide which changes to scale up.
- Build reliable evaluation sets and maintain clear data lineage so training runs can be traced and reproduced.
- Own production data pipelines and data-related problems from initial research question through shipped result.
Benefits
- Competitive base salary plus equity.
- Unlimited PTO and company-wide Refill Days, including long weekends when the whole company rests together.
- Comprehensive health coverage, with details varying by location.
- Paid leave for pregnancy, parenting, medical needs, and caring for family.
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Mid Level
Education Level
No Education Listed
