Source:
Research, Mid Training
Thinking Machines Lab
San Francisco, United States
Publication date: 06/11/2026
The mission of Thinking Machines is to build AI that extends human will and judgment. Mid-training is a step between pre-training and post-training, where a base model is trained into the foundation for reasoning. This role owns the late-stage training responsibility that shapes what the models are fundamentally capable of, including synthetic data strategies, the data mix, quality uplift, context extension, and capabilities across coding, math, reasoning, and so on. The role is an IC role within the Research Team, reporting directly to the Head of Research.
Responsibilities
Own the data: decide what the model needs to see, then source, curate, and synthesize it, and build pipelines that filter, deduplicate, verify, and rewrite raw material into training-grade datasets; Improve what the model knows: design and measure interventions that increase knowledge, and characterize how knowledge scales with data; Instill behaviors and set the prior: introduce new behaviors during mid-training and work with the post-training team on which behaviors belong in mid-training versus RL; Build the quality pipeline: own filtering and scoring, train quality classifiers and LLM judges, build verifiers for synthetic data, and control data attributes; Develop and tune the recipe: iterate on datasets, training stages, annealing schedules, and hyperparameters; Iterate on evals: define, optimize, and refine evaluations so they capture what matters; Debug and understand: diagnose unexpected training results and build deeper understanding; Scale and explore: measure how performance scales with dataset size and explore new kinds of training dataRequirements
Proficiency in Python and familiarity with deep learning frameworks (e.g., PyTorch, TensorFlow, or JAX); Comfort debugging distributed training and writing code that scales; Bachelor's degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding; Clarity in communication, with an ability to explain complex technical concepts in writingNice to have
A strong grasp of probability, statistics, and ML fundamentals; Experience building or owning training datasets for large models, including synthetic data pipelines and real user data; Experience with knowledge injection, continued pre-training, or domain adaptation of large models, and with measuring what a model knows; Experience building model-based data quality systems (quality classifiers, LLM judges, verification and rewriting pipelines); Experience generating synthetic data and understanding its failure modes (diversity collapse, contamination, teacher artifacts); Research or engineering contributions in alignment, data-centric AI, or human-AI collaborationBenefits
Generous health, dental, and vision benefits; Unlimited PTO; Paid parental leave; Relocation support as needed; Visa sponsorshipHow to apply
Apply online: https://www.jobs.pl/en/oferta/101816983/aplikuj
JOBS.pl · https://www.jobs.pl/en/research-mid-training-oferta-101816983


