Senior Research Engineer, Model Training / Pretraining

Responsibilities: Build and maintain the training stack for distributed model training across GPU clusters. Run scaling experiments covering model architecture, data, and optimisation. Work in one specialist area: GPU kernel development (CUDA), Mixture-of-Experts architectures, or reinforcement-learning post-training (GRPO, PPO, RLVR).Release models, code, and associated research publicly.

Requirements: 6+ years of software engineering experience.4+ years of machine-learning infrastructure experience. Python and PyTorch. Experience training models from scratch (pretraining). Experience limited to retrieval-augmented generation or application-level fine-tuning does not meet this requirement. Depth in at least one of: CUDA kernels, Mixture-of-Experts, or reinforcement learning (GRPO, PPO, RLVR).Also consideredJAX.Open-source contributions. Published research. Experience with GPU clusters at the scale described above.

Interview process: Hiring-manager screen — 25 minutes. Recruiter screen — 15 minutes. Machine-learning coding interviews — two sessions, 45 minutes each. Machine-learning system-design interview — 45 minutes. Technical deep-dive with the Director of AI Research — 60 minutes. Final interview — 30 minutes. Offer.

Base salary: $147,000–$220,000 + strong bonus programmes in addition to base salary. Hybrid in Seattle, Washington. On-site presence is required. Fully remote is not available. Visa sponsorship is available. Application This search is managed by Intelix. AI.

Contact: Hasan Mohammad — hasan@intelix.ai.