TL;DR: Live listings in September 2026 pay a new-grad applied AI engineer $135,000 to $155,000, a mid-level agents engineer $180,000 to $225,000, a staff engineer $252,000 to $315,000, and a prompt engineer at a frontier lab $300,000 to $405,000. Aggregators sit lower: Built In shows $211,243 total for AI engineers and Indeed shows $112,806 for prompt engineers. The skills that unlock the top bands are the same everywhere: Python, agent frameworks, MCP, and evals. Learn evals first.
The salary table: four titles, four levels, every row sourced
Aggregators average three years of postings; listings show what a hiring manager will sign this month. Figures are annual base in USD unless marked TC (total compensation), checked September 4, 2026.
| Title | Level | Figure | Source |
|---|---|---|---|
| AI engineer | Junior | $135,000 to $155,000, new grad | NewsBreak, Greenhouse |
| AI engineer | Mid | $180,000 to $225,000, 4+ years | Scale AI, Greenhouse |
| AI engineer | Mid | $200,000 to $320,000, 4+ years | Anthropic, Greenhouse |
| AI engineer | Mid | $184,757 base, $211,243 total, median $180,000 | Built In |
| AI engineer | Mid | $152,203 average, range $93,058 to $248,939 | Indeed |
| AI engineer | Senior | $216,000 to $270,000, 5+ years | Scale AI, Greenhouse |
| AI engineer | Senior | $280,000 to $320,000, 4+ years, forward deployed | Anthropic, Greenhouse |
| AI engineer | Staff | $252,000 to $315,000, 8+ years | Scale AI, Greenhouse |
| AI engineer | Staff | $288,000 to $360,000, 12+ years, senior staff | Scale AI, Greenhouse |
| ML engineer | Junior | $125,000 to $175,000, 2+ years or MS/PhD | NewsBreak, Greenhouse |
| ML engineer | Mid | $188,643 average, range $112,922 to $315,140 | Indeed |
| ML engineer | Mid | $162,080 base, $212,022 total | Built In |
| ML engineer | Mid | $248,000 average TC, ML/AI focus | Levels.fyi |
| ML engineer | Senior | $140,000 to $225,000, 3+ years | ZipRecruiter (employer), Greenhouse |
| ML engineer | Staff | $179,400 to $303,600 | XPENG, Greenhouse |
| Prompt engineer | Junior | $140,000 to $170,000, 1+ years | Cantina Labs via Built In |
| Prompt engineer | Mid | $112,806 average, range $72,377 to $175,819, 52 postings | Indeed |
| Prompt engineer | Senior/Staff | $300,000 to $405,000 (posting closed April 28, 2026) | Anthropic via Built In |
| AI product engineer | Mid | $120,000 to $220,000, agent platform | NewsBreak, Greenhouse |
| AI product engineer | Senior | $153,000 to $376,000, 5+ years | Figma, Greenhouse |
| AI product engineer | Startup average | $133,042 average, $210,375 top of market | Wellfound |
First, “AI engineer” on a Greenhouse listing pays about $45,000 more than on an aggregator. Built In’s median is $180,000; that is Scale AI’s floor for a 4-year engineer, with a $225,000 ceiling. Aggregators count every insurer’s AI engineer; AI companies’ listings do not.
Second, “prompt engineer” is two jobs. Indeed’s $112,806 average comes from 52 postings and bottoms out at $72,377. Anthropic’s Claude Code prompt engineer role paid $300,000 to $405,000 and required Python, TypeScript and behavioral evaluations. Same title, 3.6x the pay: one is a writing job, the other an engineering job.
Third, ML engineer pay is flatter. Indeed’s ML average of $188,643 beats its AI engineer figure, but ML listings ask for PyTorch, not agents, and the ceiling is lower outside the labs.
Frontier labs are a different market
Levels.fyi company pages show total compensation, base plus stock plus bonus, and stock is what doubles or triples the package.
| Company | Entry | Mid | Senior | Staff | Median TC | Source |
|---|---|---|---|---|---|---|
| Anthropic | $367K | n/a | $591K | $1.25M (lead $779K) | $850K | Levels.fyi |
| OpenAI | $250K (L2) | $333K (L3) | $652K (L4) | $942K (L5) | $943K | Levels.fyi |
| Scale AI | $231K (L3) | $370K (L4) | $476K (L5) | $1.23M (L6) | $358K | Levels.fyi |
Anthropic’s entry-level base is $250,000 with $117,000 a year in stock. Scale’s L3 base is $159,000 with $67,600 in stock, in line with Scale’s own $180,000 to $225,000 listing. Stock is the part no listing prints.
What the listings actually ask for
Requirements below are as written in the posted text.
| Listing | Pay | Languages | AI tooling named | Evals | Years |
|---|---|---|---|---|---|
| Anthropic, Forward Deployed Engineer | $280K to $320K | Python, TypeScript, Java | LLMs in production, agent development, MCP servers, sub-agents, agent skills | Evaluation frameworks | 4+ |
| Anthropic, Applied AI Engineer | $200K to $320K | Python or TypeScript | Prompt engineering, agent frameworks, MCP, transcript analysis | Evaluation frameworks | 4+ |
| Anthropic, Prompt Engineer, Claude Code | $300K to $405K | Python, TypeScript | Agentic coding tools, agent scaffold architectures | Behavioral evaluations | Level-dependent |
| Scale AI, Frontier Agents Engineer | $180K to $225K | Python | OpenAI, Claude, MCP, agent frameworks, vector databases, retrieval | Quantitative metrics | 4+ |
| Scale AI, Senior Frontier Agents Engineer | $216K to $270K | Python | Same list plus RAG, semantic search, fine-tuning, RL, AWS/Azure/GCP, Docker, Kubernetes | Quantitative metrics | 5+ |
| Indicium AI, Senior Forward Deployed AI Engineer | $180K to $240K | Python, SQL | MCP, function calling, JSON schemas, Claude, LangGraph, AutoGen, CrewAI, Databricks, PySpark, vector search, RAG | LLM evals, hallucination tracking, retrieval precision/recall | 5+ |
| NewsBreak, AI Engineer, Agent Platform | $120K to $220K | Python, Go or Node, React or Next.js | Cursor, Claude Code, Codex, v0, tool-calling, context management, MCP, sub-agents, LoRA/RLHF/DPO | LLM-as-judge, custom harnesses | Not stated |
| NewsBreak, Applied AI Engineer, New Grad | $135K to $155K | Typed code, async | LLMs, RAG, tool-use, agent orchestration, schema-validated outputs, Docker, Kubernetes | Not listed | 0 |
| Cantina Labs, Prompt Engineer | $140K to $170K | Not required | Temperature, top_p, top_k tuning, chain-of-thought, dataset creation | Evaluating model output | 1+ |
| XPENG, Staff Machine Learning Engineer | $179K to $304K | Python | PyTorch, TensorFlow, reinforcement learning, ONNX, TensorRT | Not listed | 3 to 5+ |
| ZipRecruiter, Senior Software Engineer, ML | $140K to $225K | Not specified | TensorFlow, PyTorch, scikit-learn, NLTK, SpaCy, BERT, CNN/RNN/LSTM/GAN | Not listed | 3+ |
| Figma, Software Engineer, AI Product | $153K to $376K | Not specified | GitHub Copilot, ChatGPT, Claude, retrieval, personalization, NLP | Not listed | 5+, with 1+ in genAI |
Python is required in 8 of 12 listings. MCP appears in 6, five of them with a floor of $180,000 or higher. Evaluation work appears in 8, including every listing with a floor at or above $200,000. PyTorch and TensorFlow appear only in the two ML listings, and neither tops $304,000.
Nobody asks for a model version. Listings name “OpenAI, Claude” and then MCP, agent frameworks, vector databases and evals: the stack around the model. Running evals on Claude Sonnet 5 at $2 in / $10 out per 1M tokens, or Gemini 3.8 Flash at $0.75 / $3.75, is pocket change next to the salary delta; the AI API pricing September 2026 table has every rate.
Freelance vs salaried: the hourly math
The AI coding freelance rate card lists typical public rates of $40 an hour for LLM API integration, $46 for chatbot and RAG builds, $72 for agent and workflow automation, $75 for fine-tuning and MLOps, $125 for senior ML and AI engineering with a $200 ceiling, and $200 to $240-plus for fractional AI architecture.
Salaried hourly equivalent: salary divided by 2,080 hours (40 a week, 52 weeks). Freelance break-even: salary divided by 1,200 billable hours, assuming 25 billable hours a week for 48 weeks with the rest lost to sales and admin. The 1,200 is an assumption, and generous for a first-year freelancer.
| Salary band (listing) | Per hour at 2,080 hrs | Freelance rate to match at 1,200 hrs | Rate card service that clears it |
|---|---|---|---|
| $135,000 (new grad AI engineer) | $64.90 | $112.50 | Senior ML engineering ($125 typical) |
| $180,000 (Scale mid floor) | $86.54 | $150.00 | Senior ML high band ($200) or architect |
| $225,000 (Scale mid ceiling) | $108.17 | $187.50 | Senior ML high band ($200) or architect |
| $270,000 (Scale senior ceiling) | $129.81 | $225.00 | Architect tier only ($200 to $240-plus) |
| $315,000 (Scale staff ceiling) | $151.44 | $262.50 | Above every listed tier |
| $405,000 (Anthropic prompt engineer ceiling) | $194.71 | $337.50 | Above every listed tier |
Take the $216,000 to $270,000 senior band, midpoint $243,000. At 2,080 hours that is $116.83 an hour. Matching it freelancing at 1,200 billable hours takes $202.50 an hour, before your own health insurance, self-employment tax and tooling. The rate card’s senior ML ceiling is $200. Only architect-tier work at $200 to $240-plus clears $202.50.
The other direction: an LLM API integrator at the $40 typical rate billing 1,200 hours grosses $48,000, and even the $80 top of that band grosses $96,000. Indeed’s salaried prompt engineer average is $112,806 with benefits. Entry-level freelance AI work loses at every point in the range.
The crossover: below $180,000, salaried wins unless you can bill senior ML or agent work at $125-plus from day one, which matches $150,000. Above $270,000, freelancing only matches at $225-plus, architect-tier work, where the AI consulting rate card beats hourly coding anyway.
Which title pays best per skill required
Prompt engineer at a lab wins on paper: $300,000 to $405,000 for Python, TypeScript, agent scaffolds and evals. But that posting closed in April, few like it exist, and the median prompt engineer job is a $112,806 role asking for temperature tuning and chain-of-thought. High variance, thin market.
AI engineer with agents and evals is the best risk-adjusted title. Scale AI posts a clean ladder from $180,000 to $360,000 base with one skill list at every level: Python, OpenAI and Claude APIs, MCP, agent frameworks, vector databases, quantitative evals. Anthropic pays $200,000 to $320,000 for the same list.
AI product engineer has the widest single range, $153,000 to $376,000 at Figma, because the title spans a full-stack developer with a chat feature and a lead who shipped a generative product. Vaguer skill list, vaguer pay.
ML engineer is the most skills for the least upside: TensorFlow, PyTorch, scikit-learn, NLTK, SpaCy, BERT and four network architectures for $140,000 to $225,000 at ZipRecruiter, and a $303,600 staff ceiling at XPENG for reinforcement learning plus ONNX and TensorRT. Deep, specific, capped.
BetOnAI Verdict
Learn evals first. Not prompting, not fine-tuning. Every skills-table listing with a floor of $200,000 or more names evals; the cheapest do not. Evals separate the $112,806 prompt engineer from the $405,000 one.
Do this:
-
Build one agent in Python with MCP, a vector store and a written eval harness, and put it on GitHub. That project covers the required list on six listings spanning $180,000 to $320,000 base. A million tokens each way costs $12 at Sonnet 5’s $2 in / $10 out, per the Anthropic API pricing page, or $4.50 on Gemini 3.8 Flash at $0.75 / $3.75.
-
Apply for “AI engineer” or “applied AI engineer”, not “prompt engineer” or “ML engineer”, unless you hold a PhD. Most listings, cleanest ladders, best pay per skill.
-
If you freelance, sell at the senior ML or architect tier, $125 to $240 an hour, or do not freelance. At 1,200 billable hours, $125 matches $150,000, $150 matches $180,000 and $225 matches $270,000. Below $112.50 an hour you lose to the NewsBreak new-grad listing.
Stop paying for prompt courses. Pay for tokens to run evals, and read the Levels.fyi stock line before accepting a base.
Frequently Asked Questions
How much does an AI engineer make in 2026?
Live listings checked September 4, 2026 pay $135,000 to $155,000 for a new grad, $180,000 to $225,000 at four-plus years, $216,000 to $270,000 at senior and $252,000 to $360,000 at staff level. Aggregators sit lower: Built In shows $211,243 total compensation and Indeed $152,203 average.
How much do prompt engineers earn?
It depends on which prompt engineer job. Indeed’s average across 52 postings is $112,806. Anthropic’s Claude Code prompt engineer role paid $300,000 to $405,000 and required Python, TypeScript and behavioral evaluations, which makes it an engineering job with a prompt title.
What skills do AI engineer job descriptions ask for?
Python appears in 8 of 12 listings, MCP in 6, evaluation frameworks in 8 including every listing with a $200,000-plus floor. Agent frameworks, vector databases and retrieval come next. PyTorch and TensorFlow appear only in classic ML listings, which cap lower.
Sources
- Greenhouse, NewsBreak new-grad applied AI engineer: https://job-boards.greenhouse.io/newsbreak/jobs/4700278006
- Greenhouse, NewsBreak AI engineer, agent platform: https://job-boards.greenhouse.io/newsbreak/jobs/4683873006
- Greenhouse, NewsBreak junior ML infra engineer: https://job-boards.greenhouse.io/newsbreak/jobs/4615879006
- Greenhouse, Scale AI frontier agents engineer: https://job-boards.greenhouse.io/scaleai/jobs/4720573005
- Greenhouse, Scale AI senior frontier agents engineer: https://job-boards.greenhouse.io/scaleai/jobs/4720478005
- Greenhouse, Scale AI staff frontier agents engineer: https://job-boards.greenhouse.io/scaleai/jobs/4720487005
- Greenhouse, Scale AI senior staff frontier agents engineer: https://job-boards.greenhouse.io/scaleai/jobs/4694869005
- Greenhouse, Anthropic forward deployed engineer: https://job-boards.greenhouse.io/anthropic/jobs/5302966008
- Greenhouse, Anthropic applied AI engineer: https://job-boards.greenhouse.io/anthropic/jobs/5057647008
- Built In mirror, Anthropic prompt engineer, Claude Code: https://builtin.com/job/prompt-engineer-claude-code/8826618
- Greenhouse, Indicium AI senior forward deployed AI engineer: https://job-boards.greenhouse.io/indiciumai/jobs/4944657101
- Built In, Cantina Labs prompt engineer: https://builtin.com/job/prompt-engineer/4453334
- Greenhouse, XPENG staff machine learning engineer: https://job-boards.greenhouse.io/xpengmotors/jobs/8743596002
- Greenhouse, ZipRecruiter senior software engineer, ML: https://job-boards.greenhouse.io/ziprecruiter/jobs/5167470
- Greenhouse, Figma software engineer, AI product: https://job-boards.greenhouse.io/figma/jobs/5551730004
- Built In, AI engineer salary: https://builtin.com/salaries/us/ai-engineer
- Built In, ML engineer salary: https://builtin.com/salaries/us/machine-learning-engineer
- Indeed, AI engineer salary: https://www.indeed.com/career/ai-engineer/salaries
- Indeed, ML engineer salary: https://www.indeed.com/career/machine-learning-engineer/salaries
- Indeed, prompt engineer salary: https://www.indeed.com/career/prompt-engineer/salaries
- Levels.fyi, ML/AI software engineer: https://www.levels.fyi/t/software-engineer/focus/ml-ai
- Levels.fyi, Anthropic: https://www.levels.fyi/companies/anthropic/salaries/software-engineer
- Levels.fyi, OpenAI: https://www.levels.fyi/companies/openai/salaries/software-engineer
- Levels.fyi, Scale AI: https://www.levels.fyi/companies/scale-ai/salaries/software-engineer
- Wellfound, AI startup salaries: https://wellfound.com/hiring-data/i/artificial-intelligence
- BetOnAI, AI coding freelance rates: https://betonai.net/ai-coding-is-the-highest-paying-freelance-skill-in-2026-heres-exactly-what-to-charge-real-rate-data/
- BetOnAI, AI consulting rate card: https://betonai.net/ai-consulting-rate-card-2026-what-to-charge-for-strategy-implementation-and-fractional-ai-leadership-real-rates-from-68-consultants/
- BetOnAI, AI API pricing September 2026: https://betonai.net/ai-api-pricing-september-2026/
- BetOnAI, Anthropic API pricing: https://betonai.net/anthropic-api-pricing-2026/
Written by Nik Sai
BetOnAI Editorial covers AI tools, business strategies, and technology trends. We test and review AI products hands-on, providing real revenue data and honest assessments. Follow us on X @BetOnAI_net for daily AI insights.
How we score: read the methodology
BetOnAI Editorial covers AI tools, business strategies, and technology trends. We test and review AI products hands-on, providing real revenue data and honest assessments. Follow us on X @BetOnAI_net for daily AI insights.
More from Nik Sai
