
Research Engineer, Large Language Models
Posted 2 days ago

Posted 2 days ago
This is a fully remote position, open to applicants in Canada.
• Design and conduct extensive pre-training and fine-tuning experiments, ranging from small-scale ablations to comprehensive training runs.
• Analyze scaling laws to determine model size, data composition, and hyperparameters while adhering to a fixed compute budget.
• Enhance training efficiency using parallelism techniques (data, tensor, pipeline, FSDP), mixed precision, and custom CUDA or Triton kernels.
• Identify and resolve training instabilities such as loss spikes, divergence, and hardware malfunctions.
• Develop and optimize data pipelines, which include filtering, deduplication, quality assessment, and dataset mixing.
• Apply concepts from recent research publications and evaluate their effectiveness at scale.
• Create assessments that accurately measure real model capabilities and meticulously document results.
• Collaborate closely with alignment, infrastructure, and product teams, sharing insights through well-structured write-ups.
• 3+ years of practical experience in training deep learning models, or a PhD in a relevant field (computer science, mathematics, or physics).
• Proficient in Python with extensive knowledge of PyTorch or JAX.
• Hands-on experience with Transformer architectures, including attention mechanisms, tokenization, embeddings, and positional encodings.
• Familiarity with distributed training on multi-node GPU clusters (FSDP, DeepSpeed, Megatron-LM, or similar technologies).
• Strong foundational knowledge in linear algebra, probability, statistics, and optimization techniques.
• Capability to interpret a research paper, implement it from the ground up, and evaluate its effectiveness.
• Meticulous experimental practices, including maintaining clean baselines, conducting ablations, achieving reproducible results, and tracking experiments.
• Excellent written and verbal communication skills, along with the ability to thrive in a fast-paced and ambiguous environment.
• Competitive salary and performance-based incentives.
• Opportunities for professional growth and development.
• Flexible work arrangements and a supportive team culture.
• Access to cutting-edge technologies and resources.
• Comprehensive health and wellness benefits.
Cardinal Health
BlueCross BlueShield of Tennessee
Centene Corporation
Get handpicked remote jobs straight to your inbox weekly.