Role overview
Contribute specialist machine learning expertise to the training and evaluation of large language models through paid, task-based work. Assignments involve reviewing technical explanations, code, model reasoning, and comparative outputs, with the aim of identifying errors and improving model quality. Candidates complete a short skills assessment before being invited to participate.
Responsibilities
- Check AI-generated explanations of model architectures, loss functions, and backpropagation for technical accuracy.
- Review machine learning code and notebooks, including training loops, data preparation, and evaluation work, for correctness and efficiency.
- Provide expert feedback that supports reinforcement learning from human feedback and model alignment.
- Assess model reasoning on complex technical prompts and identify where the logic fails.
- Compare model outputs using defined technical categories and performance measures.
- Review content for hallucinations, bias, logical errors, or possible training data contamination.
Requirements
- Bachelor's, master's, or doctoral degree in computer science, AI, robotics, or a related quantitative field focused on machine learning.
- Professional experience building, deploying, or fine-tuning machine learning models in production.
- Strong understanding of neural network architectures such as Transformers, CNNs, and RNNs, as well as optimization methods.
- Hands-on experience with prompt engineering, RLHF, or retrieval-augmented generation workflows.
- Advanced Python skills and expertise with PyTorch or TensorFlow/Keras; familiarity with common data and model libraries is relevant.
- Careful analytical judgment when evaluating technical claims, code, and model outputs.
Nice to have
- Experience with cloud ML platforms or MLOps tools, such as managed training services, experiment tracking, or orchestration frameworks.
- Familiarity with vector databases used in retrieval-augmented generation.
Benefits and work setup
- Flexible, remote task work with paid assignments; some tasks require an uninterrupted hour, while many are shorter.
- The source describes rates of up to $80 per hour, depending on the research task.