Prompt engineer
| Highest | $4,400 / mo · Switzerland |
|---|---|
| Experience | 0–1 yrs |
| Degree | Not required |
| First paycheck | 4 years of study |
Pay in the destination countries
Monthly, in US dollars. Each country is an Imora admission package — open one for that route's fees, language requirement and deadline.
| Country | Monthly |
|---|---|
| SwitzerlandIELTS 6.5 · CHF 0.9–8k | $4,400max |
| United StatesTOEFL 80–100 · $10–70k | $4,200 |
| NorwayIELTS 6.0–6.5 · NOK 130–260k | $3,350 |
| DenmarkIELTS 6.5 · DKK 45–135k | $2,850 |
| IrelandIELTS 6.0–6.5 · €10–25k | $2,750 |
| Hong KongIELTS 6.0–6.5 · HK–200k | $2,600 |
What does the day look like?
Your day is writing Python eval harnesses in Promptfoo, DeepEval, Ragas and LangSmith, running 500 to 5,000 test queries, reading JSON logs and confusion matrices, fixing API timeout and rate-limit errors, and versioning prompt text in Git. You draft system instructions, few-shot examples and output constraints as JSON schemas, XML wrappers or markdown tables for Claude 3.5 Sonnet, GPT-4o and Gemini 1.5 Pro, wire them to RAG stores in Pinecone, Qdrant or Chroma and chains in LangChain, LlamaIndex or provider SDKs, red-team with jailbreak and prompt-injection attacks, and measure token cost and latency to route requests to Claude 3.5 Haiku, GPT-4o-mini and Gemini 1.5 Flash.
Typing creative sentences into a chat window is under 10% of the day; the other 90% is code, tests and integration. The week is 40 hours, with 0 shift or night work. In enterprise teams there is on-call for latency anomalies, model deprecation releases or production prompt regressions.
Where does this lead?
Next step up
What do you make of this job?