About the roleWe are looking for Research engineers to help design and build safety and oversight algorithms for our AI models and products.
As a Trust and Safety Research Engineer, you will work to design and train ML models based on research progress, which detect harmful user/model behaviors and help ensure society's well-being.
You will apply your research skills to uphold our principles of safety, transparency, and oversight while enforcing our terms of service and acceptable use policies.
What you will be working on:
Design, iterate and build ML models to detect unwanted or anomalous behaviors from both users and LLM modelsWork with T&S ML engineers to review and iterate experiment ideations.
Co-author the experiment success criteria and production deployment roadmapsPartner with T&S Policy and Enforcement cross-functional teams to understand emerging and sustained abuse patterns from user prompts and behaviors.
Incorporate the insights into T&S research datasetsSurface abuse patterns to sibling research teams in the company.
Collaborate together to harden Anthropic's LLMs at the pre/post training stagesStay current with state-of-the-art research in AI and machine learning, and propose ways to apply these advancements to T&S systemsYou may be a good fit if you:Have 4+ years of experience in a research engineering or an applied research scientist position, preferably with a focus on trust and safetyHave significant Python programming experience and machine learning experienceHave proficiency in building trustworthy and safe AI technologyHave strong communication skills and ability to explain complex technical concepts to non-technical stakeholdersCare about the societal impacts and long-term implications of your work and are results orientedStrong candidates may also:Have experience fine-tuning large language models with supervised learning or reinforcement learningHave experience with machine learning frameworks like Scikit-Learn, Tensorflow, or PytorchHave experience authoring research papers in machine learning, NLP, or AI alignment or similar industry experienceHave developed evaluations for language models
#J-18808-Ljbffr