The AI job market: two distinct tracks
The AI job market is large, fast-growing, and widely misunderstood. Most coverage focuses on ML research and PhD-level roles — the hardest positions to get. Meanwhile, thousands of AI roles go unfilled every month because candidates assume they need more technical background than the role actually requires.
Technical track: Machine learning engineers, data scientists, AI researchers, MLOps engineers. These require strong math, programming, and often postgraduate education.
Non-technical track: AI product managers, AI policy and ethics roles, prompt engineers, AI sales engineers, AI trainers, AI QA specialists. These require domain expertise, communication skills, and AI fluency — not a CS degree.
Most people targeting AI jobs should start by identifying which track they're actually on.
Technical AI roles and how to break in
Machine learning engineer: The most in-demand technical AI role. Builds and deploys ML models in production. Requires Python, ML frameworks (PyTorch, TensorFlow), and experience with data pipelines. To break in: build a portfolio of end-to-end deployed projects — not just Jupyter notebooks. GitHub profile and Kaggle placements matter more than certifications alone.
Data scientist: Analyzes data to find patterns and build predictive models. Requires Python or R, SQL, and statistics. To break in: SQL plus Python fluency is the minimum bar. Build projects on publicly available datasets and specialize in a domain (finance, healthcare, marketing) to differentiate.
MLOps / AI infrastructure engineer: Manages the infrastructure that runs ML models at scale — training pipelines, model serving, monitoring, and reliability. Requires strong software engineering plus cloud platforms (AWS/GCP/Azure), Docker/Kubernetes, and familiarity with the ML lifecycle.
AI/ML researcher: Advances the state of the art in machine learning. Typically requires a PhD and a track record of published research. Entry is through academic-to-industry pipelines: publish papers, do research internships at AI labs.
Non-technical AI roles and how to break in
AI product manager: Owns the roadmap for AI-powered products. Translates between technical teams and business stakeholders. To break in: move laterally from a PM role at a company building AI products. A non-engineer ML course (fast.ai, Coursera's ML Specialization) builds enough fluency to talk credibly about model capabilities.
Prompt engineer / AI workflow designer: Designs and optimizes prompts and AI pipelines for business use cases. Requires deep familiarity with LLMs (GPT, Claude, Gemini), strong writing and reasoning skills, and basic scripting. To break in: build a public portfolio of complex prompting projects — multi-step workflows, RAG implementations, evaluation frameworks.
AI trainer / RLHF specialist: Rates and corrects AI outputs to improve model behavior. Requires strong subject-matter expertise (medical, legal, coding, etc.). Platforms like Scale AI and Surge HQ hire contract AI training work — a good entry point while building AI credentials.
AI ethics and policy roles: Works on responsible AI deployment, bias auditing, and regulatory compliance. Requires background in law, policy, philosophy, or social science combined with AI fluency. Target think tanks, NGOs, and enterprise trust and safety teams.
Skills that matter across both tracks
Python: The lingua franca of AI. Even non-technical AI roles benefit from basic scripting ability — enough to read and modify code, query APIs, and understand what engineers are doing.
Understanding of LLMs: How language models work at a conceptual level — tokenization, context windows, fine-tuning vs. prompting, hallucination — is now a baseline expectation for most AI-adjacent roles.
Data literacy: Reading and interpreting model evaluation metrics, understanding training/test splits, knowing what a confusion matrix means. Not data science — just literacy.
Domain expertise: AI is being applied everywhere. A nurse who understands AI in clinical settings, a lawyer who understands AI in legal workflows, a marketer who understands AI in campaign optimization — all are more valuable than a generalist.
The fastest path in — by background
If you're a software engineer: Add ML fundamentals (fast.ai Part 1 is excellent), build one end-to-end deployed ML project, and apply for ML engineering roles at companies *using* AI rather than *building* it. Easier entry than pure AI labs.
If you're in a non-technical field (marketing, finance, law, healthcare): Target AI roles in your domain. You're more valuable as a "healthcare professional who understands AI" than as a junior data scientist with no domain context.
If you're a student: Internships matter more than coursework. Aim for AI research internships, data science internships, or engineering roles at AI companies. Build publicly on GitHub.
If you're making a complete career change: Pick one track (technical or non-technical), go deep for 6–12 months, and document everything you build publicly. The portfolio is the credential.
For job boards: LinkedIn (filter for "AI," "machine learning," "LLM"), Wellfound/AngelList for startups, Levels.fyi for tech compensation, and AI-specific boards like AIjobs.net. For companies: Anthropic, OpenAI, Google DeepMind, Meta AI, Hugging Face, Mistral, and enterprise AI teams at Salesforce, ServiceNow, and Adobe.