Machine Learning Engineer, AllenNLP

You are a talented, hands-on engineer who thrives in a fast-paced environment, is self-directed, a team player, and knows how to get things done. You are motivated by creating real-world benefits using AI and are excited to help advance our effort to create the best-performing open large language model. You will be a part of the core team of research engineers behind post-training and aligning the OLMo (Open Language Model) through fine-tuning, instruction-tuning, and RLHF, helping us build infrastructure for the next generation of large language model research.

You will also help with data analysis and experimentation to direct the decisions we make and architectures we develop, both around modeling and data preparation. And, you are someone with experience working with open-source projects, deep knowledge of Python, and a strong understanding of modern deep learning and natural language processing.

Open Language Model (OLMo) is the AI2 LLM framework designed to provide access to data, training code, models, and evaluation code necessary to advance AI through open research to empower academics and researchers to study the science of language models collectively. The goal is to share a high quality and open language model that will provide an avenue for people in the AI research community to work directly on language models for the first time.

Following the launch of OLMo (Open Language Model), AI2 is now embarking on furthering research to provide scientific insights and solutions on how to substantially close the gap with the state-of-the-art in adapted models.

The essential functions include, but are not limited to the following:

• Building infrastructure to facilitate the next generation of LLM research.
• Optimizing training and inference for language models.
• Triaging between experiments and executing on the most impactful.
• Supporting and collaborating with an open-source community.
• Bridging the gap between cutting-edge research and a widely adopted product.
• Using software engineering best practices in a research environment.
• Releasing your contributions back to the broader community in the form of open source software, model releases, and additions to AI2’s public API and Open Research Corpus.

• Knowledge of modern deep learning and natural language processing techniques.
• Strong software engineering skills, particularly around building performant systems and debugging.
• Experience with the complete model development cycle, including data set construction, training, tuning, evaluation, performance profiling, and monitoring.
• Experience with Python and PyTorch, or Jax.
• Familiarity working with cloud compute resources (e.g. AWS) and containerization (e.g. Docker).

• Advanced degree in Data Science/CS/EE/Applied Mathematics/Statistics/ML/NLP or related fields.
• Contributions to open-source ML or research libraries (e.g. spaCy).
• Experience successfully operating models at scale in a production setting.
• Experience fine-tuning large models with tools like TRL or Open Instruct / methods like DPO, PPO, or supervised fine-tuning.
• Strong collaboration skills - our environment is small and collaborative, and we'd like you to thrive while working closely with others.

• BS or MS in Computer Science, Statistics, Engineering, Applied Mathematics, or a related quantitative field.

The physical demands described here are representative of those that must be met by a team member to successfully perform the essential functions of this position. Reasonable accommodations may be made to enable individuals with disabilities to perform the functions.

• Must be able to remain in a stationary position for long periods of time.
• The ability to communicate information and ideas so others will understand. Must be able to exchange accurate information in these situations.
• The ability to observe details at close range.
• Can work under deadlines.

The Allen Institute for Artificial Intelligence is a non-profit research institute in Seattle founded by Paul Allen. The core mission of AI2 is to contribute to humanity through…