The Market's Temperature
In 2024-2025, AI-related job postings grew 3x+ vs before. Meanwhile "can use ChatGPT" isn't differentiation, and a polarization is underway where high-rate offers concentrate on people who can build AI-using deliverables/products.
Surging Roles
AI Engineer / ML Engineer
Embed LLM/RAG/agents into business systems. Python/TypeScript and LLM API, Vector DB, evaluation-tool experience. US $180K-$500K, Japan 10-25M yen.
ML Ops / AI Ops
Model operation, evaluation pipelines, cost optimization, inference infra. Kubernetes, ML platforms, observability tools (Arize, Weights & Biases, Langfuse).
AI Product Manager
Planning/requirements/effect-measurement of AI-feature products. Demand concentrates on the PM-experience + AI-knowledge hybrid. Salary typically PM + 20-30%.
Prompt Designer / AI UX
UX design of AI features, prompt templates, feedback-loop building. Writing, UX, psychology are useful.
AI Trainer / AI Safety
RLHF/SFT data creation, model-safety evaluation, red-teaming. Scale AI, Surge AI, Anthropic, etc. expanding hiring.
What Applicants Should Emphasize
- Concrete deliverables: LLM apps on GitHub, public demos on X, tech blogs
- Evaluation experience: stories of improving accuracy/cost/latency numerically
- Failure stories: hallucination measures, cost blowups, recurrence prevention
- Toolchain: LangChain, Ollama, vLLM, various Vector DBs
- Classic basics: data structures, API design, security, cloud
Common Interview Questions
- How did you improve RAG retrieval accuracy
- What you devised for LLM cost optimization
- A case of reducing hallucination in production
- How you think about agent privilege design
- Criteria for switching to a new model
- Fine-tuning vs RAG selective use
People who can speak concrete numbers, architecture diagrams, walls faced impress more than abstract "AI can do it."
Changes in Non-Engineer Roles
- Marketers: AI content ops + brand strategy for 1.2x salary
- Sales: automate proposals/research with AI, focus on relationship-building
- Consultants: new AI-first strategy teams
- Legal: "legal who can use AI" is scarce at AI-contract-review-adopting firms
Job-Search Strategy
- Build a portfolio (implementation, blog, talks)
- Tech output on LinkedIn / X / Zenn
- Grasp salary ranges via casual interviews
- Salary negotiation: run multiple offers to raise market price
- Referrals: referral offers average 15-25% higher
Traps to Avoid
- Self-declaring "can use AI" with thin implementation experience
- Only chasing new frameworks, weak system-design basics
- Salary-focused without seeing the company's AI strategy
- Aiming for researcher roles but struggling without a PhD
Summary
The 2026 AI job market offers big opportunity to those with implementation experience + domain understanding + verbalized failures. More than role titles, how you can frame your strengths in an AI context is key. Continuous portfolio accumulation and output is the shortest route.
