Expertise
Code and Complex Reasoning
The hard part of code and reasoning tasks is not only finding people who can code. It is making tasks, environments, dependencies, validation and failure taxonomies runnable together.
Typical problems
- Code generation, repair, explanation and testing.
- Repository-level context and cross-file tasks.
- Tool use, environment interaction and complex reasoning traces.
- Expert evaluation and adjudication when automatic validation is insufficient.
Delivery focus
- Task environments and dependencies are reproducible.
- Evaluation criteria separate correct outcome, reasonable process and risk.
- Expert profiles, training, quality and hard cases are continuously updated.
- Data outcomes and model regression outcomes are accepted separately.