RoleGPSThe free map to your AI job
Updates

What changed, refresh by refresh

The map is re-read from the market and re-published. Each entry is the difference between two snapshots — a capability that moved ≥8 points, a role added, a sample that grew. Quote any line; it links to the page it came from.

04-10-2026since 03-10-2026

What moved

  • 1415 job postings now behind the map (+26) across 17 live roles.
  • AI Search Optimization Specialist — 55 postings (was 29). Rising: Implement schema and entity markup 62%→71%.
03-10-2026since 02-10-2026

What moved

Shares re-measured. The skill-to-capability mapping was tightened so a skill counts only for the capabilities that name it ("AI agents" no longer counts as MCP or tool calling; "AWS" no longer counts as a managed AI platform), and every published weight is now the measured share. The rising and falling capability shares below come from counting more strictly, not from a change in what employers ask; new roles and posting counts are real. The next refresh is measured against this one.

  • 1389 job postings now behind the map (+18) across 17 live roles (was 16).
  • New roles: AI Search Optimization Specialist.
  • AI Automation Specialist — 50 postings (was 93). Rising: Integrate LLM APIs into an application 50%→64%, Build a grounded RAG application with citations 31%→44% and Automate legacy UI tasks with RPA (UiPath / Power Automate Desktop) 12%→20%. Falling: Map a process and quantify automation ROI 57%→20%, Apply guardrails, safety and privacy controls 35%→12% and Deploy a support/sales chatbot on WhatsApp or web 27%→6%.
  • AI Content Creator — 92 postings (was 58). Rising: Edit and finish AI-generated video for social formats 68%→83%. Falling: Design and version prompts systematically 72%→42%, Communicate AI trade-offs to stakeholders 36%→7% and Explain how LLMs work and where they fail 72%→58%. No longer on the page: Build voice or vision LLM features.
  • AI-enabled Data Analyst — 82 postings (was 101). Rising: Turn analysis into a decision-ready story 60%→90%, Build dashboards that answer business questions 71%→98% and Design and read A/B tests 26%→48%. Falling: Apply responsible-AI and data-protection basics 29%→6%, Clean and transform data with Pandas 80%→68% and Map a process and quantify automation ROI 31%→21%.
  • AI Data Annotator / Labeling QA — 68 postings (was 105). Falling: Work in annotation tools (Label Studio / CVAT / Labelbox) 71%→13%, Apply responsible-AI and data-protection basics 32%→4% and Explain how LLMs work and where they fail 38%→15%.
  • AI Engineer — 161 postings (was 122). Rising: Build and consume REST APIs 44%→73%, Build an LLM evaluation harness 22%→46% and Build a grounded RAG application with citations 52%→75%. Falling: Integrate LLM APIs into an application 91%→47%.
  • AI Governance & Compliance Analyst. Rising: Translate business requirements into an AI solution design 32%→47%. Falling: Write labeling guidelines and evaluation rubrics 59%→0%, Explain how LLMs work and where they fail 77%→32% and Apply guardrails, safety and privacy controls 68%→36%. No longer on the page: Deliver technical demos and answer RFPs.
  • AI Product Manager — 68 postings (was 105). Rising: Build a multi-step agent workflow 41%→63% and Design and read A/B tests 24%→44%. Falling: Communicate AI trade-offs to stakeholders 82%→51%, Apply responsible-AI and data-protection basics 41%→22% and Query and model data with SQL 26%→10%.
  • AI Quality Assurance / GenAI Test Engineer — 91 postings (was 88). Rising: Write production-quality Python for AI work 67%→100%, Train and evaluate classical ML models 40%→65% and Build and consume REST APIs 60%→71%. Falling: Write labeling guidelines and evaluation rubrics 47%→5%, Evaluate and harden agents 27%→2% and Explain how LLMs work and where they fail 93%→79%.
  • AI Security Engineer — 92 postings (was 90). Rising: Secure keys, auth and data in AI apps 59%→98%, Integrate LLM APIs into an application 32%→43% and Write production-quality Python for AI work 68%→78%. Falling: Evaluate and harden agents 64%→0%, Build an LLM evaluation harness 73%→17% and Trace, monitor and debug LLM apps in production 55%→7%.
  • AI Solutions / Pre-sales Engineer — 76 postings (was 102). Rising: Run a customer discovery → POC → pilot loop 77%→88% and Translate business requirements into an AI solution design 65%→75%. Falling: Use managed AI platforms (Bedrock / Vertex / Azure AI Foundry) 55%→13%, Deliver technical demos and answer RFPs 68%→42% and Apply responsible-AI and data-protection basics 36%→20%.
  • AI Support Engineer — 50 postings (was 26). Rising: Deploy an AI service to the cloud 62%→72%. Falling: Trace, monitor and debug LLM apps in production 69%→22%, Communicate AI trade-offs to stakeholders 88%→62% and Containerize an application with Docker 42%→18%.
  • AI Technical Trainer / Instructor — 54 postings (was 28). Rising: Communicate AI trade-offs to stakeholders 36%→69% and Build dashboards that answer business questions 29%→46%. Falling: Design and deliver AI training sessions 100%→85% and Explain how LLMs work and where they fail 71%→61%.
  • Computer Vision Engineer — 83 postings (was 56). Falling: Optimize inference cost and latency 69%→33%, Work in annotation tools (Label Studio / CVAT / Labelbox) 27%→8% and Integrate LLM APIs into an application 19%→5%.
  • Forward Deployed Engineer (AI) — 104 postings (was 101). Rising: Translate business requirements into an AI solution design 49%→68%, Build a multi-step agent workflow 51%→67% and Design scalable AI-backed systems 46%→55%. Falling: Integrate LLM APIs into an application 57%→34%, Deliver technical demos and answer RFPs 20%→8% and Build voice or vision LLM features 23%→12%.
  • Machine Learning Engineer — 88 postings (was 103). Rising: Train and evaluate classical ML models 77%→99% and Build scheduled data pipelines 45%→60%. Falling: Integrate LLM APIs into an application 45%→25%, Deploy an AI service to the cloud 77%→62% and Apply responsible-AI and data-protection basics 26%→11%.
  • Prompt Engineer / AI Workflow Specialist — 113 postings (was 105). Rising: Build an end-to-end chat assistant 44%→73%, Write production-quality Python for AI work 47%→66% and Build a grounded RAG application with citations 28%→42%. Falling: Build an LLM evaluation harness 69%→40%, Apply guardrails, safety and privacy controls 44%→26% and Build voice or vision LLM features 38%→20%.
02-10-2026since 17-09-2026

What moved

17-09-2026since 09-09-2026

What moved

09-09-2026since 04-09-2026

What moved

04-09-2026since 30-08-2026

What moved

  • Retired: GenAI / LLM Application Engineer (folded into AI Engineer; its postings are counted there now).
  • 371 job postings now behind the map across 12 live roles (was 13).
  • AI Engineer — 54 postings (was 29). Falling: Implement tool / function calling 24%→15%, Write production-quality Python for AI work 69%→61% and Build and consume REST APIs 52%→44%.
  • AI Solutions / Pre-sales Engineer. Rising: Train and evaluate classical ML models 23%→32%. Newly asked for: Apply responsible-AI and data-protection basics (23%). No longer on the page: Build scheduled data pipelines and Build a grounded RAG application with citations.
  • Forward Deployed Engineer (AI). Falling: Deliver technical demos and answer RFPs 40%→20%, Translate business requirements into an AI solution design 60%→49% and Communicate AI trade-offs to stakeholders 80%→71%.
30-08-2026since 24-08-2026

What moved

  • 371 job postings now behind the map (+146) across 13 live roles (was 10).
  • New roles: AI Governance & Compliance Analyst, AI Quality Assurance / GenAI Test Engineer, AI Security Engineer.
  • AI-enabled Data Analyst — 35 postings (was 22). Rising: Build scheduled data pipelines 27%→37%. Falling: Write production-quality Python for AI work 64%→51%, Build dashboards that answer business questions 82%→71% and Turn analysis into a decision-ready story 68%→60%.
  • AI Data Annotator / Labeling QA — 34 postings (was 22). Falling: Evaluate and rank model responses (RLHF / preference data) 59%→41% and Explain how LLMs work and where they fail 50%→38%.
  • AI Engineer — 29 postings (was 22). Falling: Train and evaluate classical ML models 36%→28%.
  • AI Product Manager — 34 postings (was 23). Falling: Prototype an AI feature without an engineering team 48%→35%.
  • AI Solutions / Pre-sales Engineer — 31 postings (was 21). Rising: Design scalable AI-backed systems 43%→61%, Translate business requirements into an AI solution design 57%→71% and Explain how LLMs work and where they fail 62%→74%. Falling: Build scheduled data pipelines 29%→19%.
  • Forward Deployed Engineer (AI) — 35 postings (was 23). Rising: Integrate LLM APIs into an application 52%→66%, Design scalable AI-backed systems 35%→49% and Translate business requirements into an AI solution design 48%→60%.
  • Machine Learning Engineer — 31 postings (was 20). Rising: Trace, monitor and debug LLM apps in production 15%→29% and Integrate LLM APIs into an application 35%→45%. Falling: Build image models (detection/classification) 40%→26%. Newly asked for: Apply responsible-AI and data-protection basics (26%) and Build a multi-step agent workflow (23%).
  • Prompt Engineer / AI Workflow Specialist — 32 postings (was 21). Rising: Build voice or vision LLM features 29%→38%. Falling: Get reliable structured outputs from LLMs 29%→19%.
24-08-2026baseline

The first map

  • 10 roles mapped from 225 job postings read in full: AI Automation Specialist, AI-enabled Data Analyst, AI Data Annotator / Labeling QA, AI Engineer, AI Product Manager, AI Solutions / Pre-sales Engineer, Forward Deployed Engineer (AI), GenAI / LLM Application Engineer, Machine Learning Engineer and Prompt Engineer / AI Workflow Specialist.
  • 57 capabilities with a measured share of job postings; 209 free, link-checked resources.
  • This is the snapshot every later refresh is measured against.

Threshold and counting rules are on the methodology page; the full table is open data. Partners who see drift before the postings do can tell us.