Describe What You Do In AI Terms
Turn 'I made the chatbot work' into language that survives a CV, an interview, and a technical screening.
Most developers are doing more AI engineering than their CV says. The problem is vocabulary. "I used ChatGPT in the app" sounds disposable. "Implemented retrieval-augmented generation with structured outputs and evaluation loops" sounds like a thing a company can pay for.
Do not lie. Translate.
If you built a chatbot over company docs
Say: Built a retrieval-augmented generation pipeline over internal documents.
Useful details:
- chunking and indexing documents
- embedding search or hybrid search
- prompt construction with retrieved context
- citation or source display
- fallback behavior when retrieval is weak
If the model calls APIs
Say: Implemented tool use / function calling for external actions.
Useful details:
- JSON schema for tool arguments
- server-side validation
- permission checks before actions
- retry and error handling
- audit logs for tool calls
If the output must be machine-readable
Say: Used structured outputs to make LLM responses reliable for downstream code.
Useful details:
- schema-constrained output
- parsing and validation
- repair paths for invalid responses
- typed application integration
If you checked whether it works
Say: Built evals for LLM behavior.
Useful details:
- golden test cases
- regression checks
- human review labels
- pass/fail criteria
- trace review
The CV shape
Use this pattern:
Built [AI capability] using [model/provider/framework], with [retrieval/tool use/structured output], measured by [evals/observability], improving [business or developer outcome].
That sentence is not magic. It is just the difference between "I used AI" and "I can explain the system."