Scale’s mission is to accelerate the development of AI and AI-powered applications. With a focus in human labeling and annotation, Scale services are trusted by industry-leading businesses such as OpenAI, Lyft, Meta, GM, Samsung, Airbnb, Microsoft, Meta, and NVIDIA. Scale works at the frontier of AI development, trust, and safety. Our workflows include development of LLM applications for commercial purposes, improvement of training data sets via model capabilities, and discovery of new evaluations to make models more trustworthy and reliable. We’re seeking creative problem solvers, ML researchers, and red team hackers to help us push the frontier of the next generation of AI model capabilities for both performance and safety.

You will:

• Develop LLM applications to evaluate data quality, solve difficult problems of synthetic data capabilities and execute business solutions
• “Jailbreak” and find vulnerabilities in LLMs and other generative AI models. Develop training materials for red teaming and adversarial analysis; conduct training with operation staff
• Develop data creation and vulnerability testing scripts in Python/TypeScript; work with ML Engineers and software engineers for robust implementation and in-platform deployment
• Analyze and visualize experimental prompt results
• Develop and teach new prompt engineering methods to ML engineers and software engineering staff

You Have:

• B.S. or higher in a quantitative major or one related to language models (linguistics, philosophy, etc.)
• Proficiency writing Python or TypeScript for API-enabled scripts
• Ability to conduct analysis of model performance via statistics and data visualization
• Experience using frontier-scale LLMs (GPT, Claude, PaLM) programmatically towards specific applications or implementations
• Demonstrated ability to have built an LLM-based application or hacking an LLM to serve some novel purpose
• Familiarity with LLM evaluation metrics (BLUE, ROUGE, MAUVE, etc.) and statistics
• Strong written and verbal communication skills
• A creative mindset to drive experimentation and internal R&D

Nice to haves:

• Previous experience with AI/LLM developmental work or red teaming a strong plus
• Academic work in language models, particular on adversarial robustness
• Previous disclosures of vulnerabilities or security issues in LLMs or other AI models
• Experience working with cloud technology stack (eg. AWS or GCP) and developing machine learning models in a cloud environment