Member of Technical Staff, Pretraining Data
Mirendil · San Francisco
Frontier labonsite$300k–$400k
At a glance
Highlights
- Set direction for what frontier models train on and prove it with experiments.
- Scale pretraining corpus to larger models across code and science domains.
- Keep growing corpus clean with provenance and reproducible versions.
What they offer
- meaningful equity grant
- competitive benefits
Ideal candidate
You are a judgment-driven researcher-engineer who has shaped large pretraining runs and cares about what models learn. You reason clearly from experiments and love directing agents to test data ideas at scale.
Requirements
- Shipped a data change into a large pretraining run with ablation to justify it
- Built data pipelines at trillions-of-tokens scale with Spark, Ray, Beam or similar and strong Python
- Know PyTorch or JAX well enough to run own training experiments
- Hands-on experience with at least two of: deduplication, quality classifiers, mixture optimization, data scaling laws, or long-context and code data
- Can tell when an experimental result is real and explain reasoning clearly in writing
Lab profile
About Mirendil
Frontier AI research company training its own models to accelerate science and technology.
Known for
- End-to-end frontier model training
- Pretraining and custom ML kernels
- RL post-training and environments
- Evaluation, inference and recursive self-improvement
Posted · Last seen