Data Annotation Specialist, Data Science

Cohere Cohere · AI Frontier · Canada · Data Quality (Contract)

This role focuses on evaluating data science and coding tasks, requiring review and debugging of code, analysis of model trajectories, and assessment of data visualization script development and logical flow implementation. The work contributes to model development efforts and the logic models apply when completing task requests. The role involves prompting models for data science tasks, labeling and improving outputs, and reporting on model/agent behavior. Familiarity with code agents or evaluating agent trajectories is a plus.

What you'd actually do

  1. Evaluate the model's ability to respond to coding requests, workflows, and code base-related questions using available tools.
  2. Assess agent trajectories and model capabilities for code generation, tabular and graphic manipulation, and debugging requests.
  3. Prompt models to complete complex data science tasks and review the accuracy of generated responses.
  4. Label, proofread, and improve machine-written and human-written software engineering-related outputs.
  5. Report quality and performance trends related to model/agent behaviour and project assignments.

Skills

Required

  • Exceptional data analysis and visualization skills
  • 3+ years of relevant industry experience working on real-world data science problems and pipelines
  • Proficient knowledge and understanding of Python and industry-standard data science packages (numpy, pandas, matplotlib, sqlite, or others)
  • Strong understanding of SQL syntax writing and workflows
  • Deep familiarity with file/data formats, such as markdown, JSON, XML, YAML, and HTML
  • Prior experience re-writing, proofreading, and delivering feedback on code

Nice to have

  • Familiarity with code agents (OpenCode, Claude Code, Codex) or evaluating agent trajectories

What the JD emphasized

  • review and debug code
  • analyze model trajectories
  • assess data visualization script development
  • evaluate the model's ability to respond to coding requests
  • assess agent trajectories and model capabilities for code generation
  • Prompt models to complete complex data science tasks
  • Label, proofread, and improve machine-written and human-written software engineering-related outputs
  • Report quality and performance trends related to model/agent behaviour

Other signals

  • evaluating data science and coding tasks
  • review and debug code
  • analyze model trajectories
  • assess data visualization script development
  • evaluate the model's ability to respond to coding requests
  • assess agent trajectories and model capabilities for code generation
  • Prompt models to complete complex data science tasks
  • Label, proofread, and improve machine-written and human-written software engineering-related outputs
  • Report quality and performance trends related to model/agent behaviour