Remote LLM Research Scientist (Pre/Post-Training)

Synthires; California; Remote; USD 138,000 - 165,000; Part time

Job summary

Synthires is seeking an LLM Research Scientist for remote, hourly engagement to advance state-of-the-art foundation models. You will work on pre-training and post-training, data curation, model evaluation, and optimization across challenging research problems.

You will design pipelines for pre-training, fine-tuning, SFT, and reward modeling, while collaborating with AI researchers to improve reasoning, efficiency, and alignment in a remote setting.

Qualifications

• 3+ years of machine learning research experience (PhD research qualifies).
• Strong expertise in LLM post-training, pre-training, evaluation, and alignment.
• Experience with RLHF, DPO, or related preference optimization methods.

Responsibilities

• Train transformer-based language models from scratch and fine-tune open-weight foundation models.
• Design and optimize pre-training and post-training pipelines for large language models.
• Construct, curate, and optimize large-scale training datasets from raw web data sources.
• Develop data filtering, deduplication, and curriculum learning strategies.
• Diagnose and resolve training instabilities and convergence issues.
• Build and evaluate SFT, RLHF/RLAIF, and reward modeling pipelines.
• Design evaluation benchmarks and analyze model performance.
• Collaborate with researchers to improve model reasoning and alignment.

Skills

Machine learning research

PyTorch

JAX

TensorFlow

Transformer architectures

Model evaluation

RLHF / RLAIF

Python

Research communication

Education

PhD in CS / ML / AI

Tools

PyTorch

JAX

TensorFlow

Job description

Synthires is seeking an LLM Research Scientist for remote, hourly engagement to advance state-of-the-art foundation models. You will work on pre-training and post-training, data curation, model evaluation, and optimization across challenging research problems.

You will design pipelines for pre-training, fine-tuning, SFT, and reward modeling, while collaborating with AI researchers to improve reasoning, efficiency, and alignment in a remote setting.