Career guide · Reviewed 9 September 2026
Can you work in AI evals without coding?
Yes, some evaluation work does not require software engineering. The opportunity depends on the task, your subject knowledge and the employer's requirements.
“Evals” describes a kind of work, not one standardized job title. You may see it inside human data, quality operations, AI training, research or engineering roles. A nontechnical title does not guarantee that no coding or analytics is needed.
Three routes into the work
| Route | What you contribute | What to check |
|---|---|---|
| Domain expert or human evaluator | Realistic tasks, accurate judgments, examples and rubrics. | Required qualifications, experience, language and whether any coding is specified. |
| Evaluation or human-data operations | Clear instructions, reviewer calibration, quality checks and project delivery. | Analytics requirements, technical tooling and previous operations experience. |
| Evaluation engineer or researcher | Automated tests, environments, graders and experiments. | Programming, statistics and research requirements; this is usually a technical route. |
What actual employers ask for
Direct employment: OpenAI Human Data Program Manager
The listing includes defining success criteria, writing instructions and calibrating AI trainers. It asks for hands-on data analytics experience and ideally one to two years of relevant experience. Software engineering is not listed as a requirement. This is an operations role supporting training and evaluation, not an entry-level evaluator job.
Read the official role and current requirements
Project contribution: experienced domain experts
OpenAI's knowledge-work interest form seeks professionals with at least five years of industry experience for vendor-facilitated projects. Completing it is an expression of interest for future participation, not a job offer or a confirmed assignment.
Examples checked 9 September 2026. Availability can change. Contributing through a vendor is different from being employed by the AI lab.
How your background could transfer
These are suggested practice directions, not claims that every employer hires each background.
| Your background | A useful practice task | Evidence to show |
|---|---|---|
| Writing and editing | Check a summary against its source. | A rubric separating factual accuracy, omissions and writing quality. |
| Teaching and education | Assess a tutor's explanation of a familiar topic. | Examples distinguishing a correct answer from a useful explanation. |
| Customer support and operations | Check whether an agent follows a fictional refund policy. | Pass/fail decisions with evidence and escalation rules. |
| Professional domain expertise | Assess an answer within your field using appropriate reference material. | Documented expertise, task boundaries and reasons for each judgment. |
Read the opportunity carefully
- Employment: check the employer, location, contract type and benefits.
- Freelance or contributor work: check the contracting entity, task availability and payment terms. An advertised hourly maximum is not guaranteed income.
- Experience: “no coding” does not mean “no expertise.” Some projects need specialist credentials or several years in a profession.
- Technical requirements: look for Python, SQL, APIs and statistical analysis in the full description. If unspecified, treat them as unknown.
Make your judgment visible
Start with a small portfolio: a clear task, a rubric, scored examples and a short explanation of disagreements and limitations. Show how you reached the decision, including cases where you were unsure.
Try the first evaluation project →
Browse human-data roles — this category includes varied technical requirements; read each listing.