Prime Intellect
Research Engineer - Pre-training
$350K
San Francisco +1 · Full-time · Posted Oct 16, 2026
- Location
- San Francisco
Remote - Experience
- No specific number of years stated.
- Visa sponsorship
- Visa sponsorship and relocation assistance for international candidates.
About the job
Build and improve pre-training systems and run experiments for Prime Intellect’s open language models, scaling toward 1–3T parameters.
What you'll do
- Extend pre-training code and take experiments into full training runs; evaluate architecture changes against model quality and training cost.
- Improve training throughput, reliability, checkpointing, and recovery; investigate training dynamics and numerical stability.
- Run reproducible ablations and collaborate on training mixtures, midtraining, and long-context experiments.
What they ask for
- Hands-on language-model training experience, including modifying training code and evaluating actual runs.
- Strong Python skills and experience with PyTorch or a comparable framework, plus systems understanding to debug distributed training.
- Practical understanding of model architectures and optimization, including effects on convergence.
- Ability to design and interpret research ablations.
Pay
$150-350k (cash compensation range, including equity incentives; currency not stated)
Tools
prime-rl, Python, PyTorch, MoE
AI focus
Pre-train open language models; evaluate model architectures and training methods, including experiments scaling toward 1–3T parameters.
Model-written summary · the original posting is authoritative