Description
We are committed to the development of artificial intelligence and understand that creating a safe and reliable platform is crucial for our mission and our customers. Therefore, we are looking for an engineer to help us develop capabilities in the field of trust and safety.
In this role, you will work on systems to prevent risks related to factual and contextual hallucinations in large language models and AI agents, manifestations of toxicity and discrimination, and the prevention of abuse. You will train and fine-tune multimodal models, create new model architectures, and design tactics for their application, such as creating guardrails.
If you are inspired by the spirit of Bell Labs/Xerox PARC in their best years — to explore, test hypotheses, and see them through to launch — then we are the place for you.
Responsibilities
- develop, build, and maintain solutions to combat abuse, moderate content, and detect hallucinations, designed to protect the company and end users from undesirable behavior
- closely collaborate with other engineers and researchers to utilize both standard and novel AI methods for measuring, monitoring, and improving the alignment of AI models and AI agents with human values
- study papers and materials from relevant Russian and international conferences on AI Safety and participate in them
- identify and resolve active incidents on the platform, as well as create new tools and infrastructure that address the root causes of system failures.
Requirements
- experience in the field of content safety, fraud or abuse prevention, OR are motivated and eager to work on current AI safety issues
- proficiency in Python or modern languages such as C++, Rust, or Go, and willingness to quickly master Python
- experience deploying classifiers or machine learning models, or a desire to get acquainted with modern ML infrastructure
- experience designing and training neural network architectures — from classic CNN/RNN to modern transformers
- experience in developing AI agents and a general understanding of advanced MAS architectures
- experience creating and supporting production services in conditions of rapid growth and scaling
- ability to critically assess the risks of a new product or feature and find innovative solutions for their mitigation without degrading the user experience.
Conditions
- comfortable, modern office near Kutuzovskaya metro station
- hybrid work format
- annual salary review and annual bonus
- corporate gym and recreation areas
- access to over 400 educational programs from SberUniversity for professional and career development
- extended voluntary health insurance, preferential insurance for family members
- flexible mortgage discount equal to 1/3 of the Central Bank's key rate
- free subscription to SberPrime+, discounts on products from partner companies
- referral bonus for recommending friends to join Sber's team.
