Weekday AI · posted · checked
Software Engineer – Backend
C++JavaPythonRustTypeScriptRemote · United States
Who can apply
Remote · United States
Location: United States
This role is for one of our clients
Compensation: $50–$100 per hour
Engagement Type: Project-based contractor role, with payment per approved task
Time Commitment: Approximately 15 hours per week
Estimated Task Duration: Around 5 hours for trainers and 1.5 hours for reviewers
Work Arrangement: Fully remote
Technical Requirements: Hands-on expertise in at least one of Python, Java, Rust, Go, C++, or TypeScript
We are looking for experienced Backend Software Engineers to contribute to the development of Reinforcement Learning Environments (RLEs) designed to evaluate and improve the capabilities of advanced AI models in solving real-world software engineering challenges.
In this role, you will create complex, repository-based coding tasks that assess a model's ability to debug sophisticated issues, implement backend features, refactor existing systems, and optimize application performance. You will work with real codebases to define engineering problems, establish reproducible development environments, and develop reliable verification systems that evaluate submitted solutions.
The ideal candidate has strong backend engineering experience, excellent problem-solving abilities, and the technical depth to identify meaningful challenges within complex software systems. Prior experience in AI or machine learning is not mandatory.
Requirements
Key Responsibilities
Design realistic and challenging backend engineering tasks using repositories and codebases with which you have hands-on experience.
Develop problems involving complex bug fixes, backend feature implementation, API development, system refactoring, and performance optimization.
Show the full posting
Build reproducible environments that allow coding tasks to be executed and evaluated consistently.
Create deterministic verifiers that identify correct solutions across different valid implementations while rejecting incomplete, incorrect, or hard-coded approaches.