135 Snow Planet Jobs

What you get to do in this role:

The CoreLLM team at ServiceNow leads groundbreaking research, engineering, and flawless execution, advancing in the key phases of Large Language Models including data curation, training, and evaluation. Our goal is to consistently enhance and improve ServiceNows signature model, NOWLLM, by incorporating innovative features and optimising efficiency, establishing new benchmarks for large-scale AI systems.

This model (NOWLLM) is consumed across various Business Units (BUs) within ServiceNow to power a wide range of BU-specific use cases, enabling teams to solve complex problems and deliver tailored solutions that accelerate innovation and efficiency across the platform.

About the Role
A crucial element in advancing Large Language Models (LLMs) is developing and leveraging high-quality datasets. Synthetic data has become a promising avenue for creating such datasets. We are looking for Research Engineers/Scientists to help generate use case specific synthetic data, conduct research on its strengths and limitations, and use it to train the cutting-edge models.

The ideal candidates will possess a strong technical foundation and a comprehensive understanding of LLM frameworks. Hands on experience in applied machine learning is required, with familiarity in synthetic data being advantageous but not essential. We are seeking individuals who excel in problem-solving, whether its crafting innovative solutions, analysing datasets, refining baselines, designing evaluations, or performing detailed scientific research.

Key Responsibilities
Develop and generate various types of synthetic data for model training.
Conducting experiments to assess the quality of synthetic data.
Train models using the curated dataset.
Design comprehensive evaluation metrics to measure the effectiveness models performance, including accuracy, relevance, generalisation ability, and robustness.
Hands on experience in hosting and optimising LLM models on Triton Inference Server, ensuring efficient model deployment while meeting performance standards and business requirements.