AI Safety Research Engineer
Posted
Summary
Builds automated evaluation infrastructure for frontier large models and AI agent systems: integrating models, benchmarks and harnesses, productizing AI safety evaluation/risk/monitoring tech, and deploying safety services for AI developers. Core stack: Python, Docker/Kubernetes, Linux, Git, and large-model APIs.
You will build automated evaluation infrastructure for frontier large models and AI agent systems. You will integrate models, benchmarks, and harnesses; develop safety products; optimize platform performance and reliability; and deploy safety services for client AI systems.
Responsibilities
- Build automated evaluation platforms for large models, multimodal models, harnesses, AI agents, and multi-agent systems
- Develop model integration, benchmark integration, harness integration, request routing, task scheduling, error handling, cost optimization, and data analysis capabilities
- Integrate open-source and in-house benchmarks
- Ensure platform feature completeness, result accuracy, and operational stability
- Productize AI safety evaluation, risk modeling, alignment, and monitoring technologies
- Optimize product functionality, performance, usability, and stability
- Deploy safety products as services for large-model and AI-agent developers
- Provide implementation plans, evaluation reports, vulnerability assessments, and improvement recommendations
Requirements
- Bachelor’s degree or above
- Python
- Data structures and algorithms
- Linux CLI
- Git
- Docker
- Kubernetes
- Large-model API integration
- Model deployment
- Benchmarking
- Technical documentation
- Chinese reading, writing, and spoken communication