What You’ll Learn
If you’ve asked “What is the salary of a prompt engineer?” you’ve probably heard a frustrating answer: it depends. Titles vary, the work ranges from UX‑like prompt crafting to full‑stack LLM engineering, and pay swings with location, company stage, and your ability to ship value with AI. Viral posts about $300k–$500k roles do exist, but most offers cluster in a more predictable band after you normalize for those factors.
In this tutorial, you’ll learn a repeatable process to benchmark prompt‑engineer (and adjacent LLM engineer) salaries in 7 steps. You’ll gather trustworthy data, compare apples to apples across locations and titles, and turn your research into a confident negotiation plan.
As a baseline for context, here’s where broader software pay sits in the U.S. (useful when a job blends software + AI responsibilities):
Use that as a floor reference; prompt/LLM roles at top tech and high‑growth AI companies often exceed it, sometimes significantly, especially when equity is meaningful.
Before You Start: Tools and Setup
You’ll move faster with a small toolkit:
- A simple spreadsheet (columns: Title, Company, Location, Level, Base, Bonus, Equity, Total Comp, Source, Notes)
- Accounts at a few data sites (read‑only is fine): Levels.fyi, Glassdoor, LinkedIn Salary, Indeed, Payscale
- A rough cost‑of‑labor reference for geo adjustments (see each site’s location filters)
- A shortlist of target companies and locations (hybrid/remote rules matter)
TipOne spreadsheet to rule them all
Create a single "comp tracker" spreadsheet and paste every data point with the date and a source link. Tidy inputs now = faster decisions later.
Where to find reliable compensation data
| Source | What you get | Good for |
|---|---|---|
| Levels.fyi | Crowdsourced base/bonus/equity and levels | Big‑tech and late‑stage calibration |
| Glassdoor Salaries | Company‑reported ranges and employee estimates | Broad coverage across industries |
| LinkedIn Salary | Aggregated ranges by title and region | Quick geo comparisons |
| Indeed Salaries | Role‑based pay snapshots | Rapid scans by location |
| Payscale | Market ranges by skills/experience | Skill premiums |
| Hired State of Salaries | Annual tech salary trends | Macro context |
The Steps
-
Define the role (and its aliases)
"Prompt engineer" can mean anything from crafting system prompts and evaluation harnesses to building full RAG pipelines. Read 10–15 current JDs across your target companies. Note must‑have skills (Python, LLM APIs, evaluation, vector databases, LangChain/LlamaIndex, RLHF exposure, safety/guardrails). Also record alias titles: "LLM Engineer," "Applied AI Engineer," "AI Engineer," "Prompt & Evaluation Engineer." Search and benchmark all relevant titles, not just "prompt engineer." For headline outliers (e.g., Anthropic’s widely cited listing), treat them as top‑end signals, not the norm (The Verge). -
Build your target market list
Make a list of 10–20 companies that match your risk/return profile: frontier labs, big tech, AI‑first startups, and AI‑curious incumbents. Capture: team size, funding stage, geo, remote policy, and whether the role sits in product, platform, or research. Pay varies across these axes—frontier labs may pay high cash or high equity, incumbents often have structured salary bands, and early startups trade more equity for cash. -
Pull fresh data from multiple sources
Search by title + location on Levels.fyi, Glassdoor, LinkedIn Salary, Indeed, and Payscale. Pull base, bonus, equity, and (if available) total compensation (TC). For roles with sparse data, triangulate from adjacent titles (e.g., "Machine Learning Engineer" or "AI Engineer") at the same company and level. Save the raw link and the date—you’ll need it to defend your range later. On company sites, look for salary ranges in job posts (many U.S. states require posting comp ranges).
-
Normalize to compare apples‑to‑apples
Not all dollars are equal. Convert every data point into an annualized total comp: TC = Base + Bonus (target) + Equity (annualized at current grant value). If equity vests over 4 years with a 1‑year cliff, divide the current fair value by 4. Factor any signing bonus separately (it’s one‑time). For geo differences, check location adjustments on sources or use each site’s city filter to see local medians.- Quick formula: Annualized equity = (Total Equity Grant ÷ 4).
- Adjust for remote: Some companies pay by employee location; others pay by office location. Ask directly which policy applies.
-
Weigh skills and impact signals
Skills shift pay. Demonstrable experience with LLM evaluation, retrieval‑augmented generation (RAG), productionizing prompts, safety/guardrails, and cost/latency optimization can place you in higher bands. Certifications are nice; shipped outcomes are better. Add portfolio signals: a small repo with evals, a fast demo app, or clear cost benchmarks (tokens, latency) can justify the top half of your range. -
Craft your target range and negotiation anchor
After normalization, compute a range (e.g., 50th–75th percentile) for your target title, level, and location. That becomes your “fair market” story. Keep two numbers handy: your confident anchor (top, defensible) and your walk‑away floor. Use your portfolio and impact narrative to support the anchor.
- Validate with real humans
Sanity‑check your numbers with trusted peers or mentors (Slack/Discord communities, alumni channels, or AI engineering meetups). Ask for ranges, not exact numbers, and never pressure others to disclose. If you have a recruiter you trust, politely ask, "Where does this role tend to land for someone with my experience?" Combine that feedback with your spreadsheet to finalize your anchor and floor.
Pro Tips & Common Mistakes
- Ask companies about their pay policy for remote (employee‑location vs office‑location). It can swing cash by 10–20%.
- Always compare total compensation, not just base. Equity and bonus often decide outcomes.
- Use multiple titles when searching ("Prompt Engineer," "LLM Engineer," "Applied AI Engineer").
- Keep screenshots or links for every data point; it’s persuasive during negotiations.
- Calibrate expectations by company stage: startups may pay below cash market but above in equity.
- When asked for expectations, reply with a researched range anchored to TC: "Based on market data for this level and location, I’m targeting $A–$B in total compensation."
Wrap‑Up
There isn’t a single “prompt engineer salary.” There’s your role definition, skills, location, company stage—and a market range you can defend. By pulling multiple sources, normalizing for geo and equity, and packaging your value, you’ll move from guessing to negotiating with conviction. Build your comp tracker today, gather five solid benchmarks, and set an anchor you can say out loud without flinching. You’ve got this.
