As of September 2026, demand for AI skills remains strong: Several reports state that jobs requiring specific AI skills have been growing substantially faster than the overall job market, while current posting analyses show especially strong demand around AI engineering, machine learning, LLMs/agents, MLOps, data infrastructure, and robotics.
For someone approaching this from an engineering/software/testing background, I’d break the market into these major career paths.
1. AI Engineer / Generative AI Engineer
This is one of the broadest and most useful AI engineering roles right now. Instead of inventing a foundation model from scratch, AI engineers usually build applications and systems around existing models.
Core skills: Python; REST APIs; software engineering; Git; SQL; cloud platforms such as AWS/Azure/GCP; Docker; CI/CD.
AI skills: LLMs, prompt engineering, RAG, embeddings/vector databases, AI agents, tool/function calling, model APIs, fine-tuning basics, guardrails, AI evaluation and observability.
Python appears in about 59% of one August 2026 sample of AI Engineer postings, followed by LLMs at 45%, ML at 29.6%, RAG at 27.3%, and prompt engineering at 18.7%.
Typical stack: Python → API → LLM → RAG → Vector DB → Agent → Evaluation → Cloud
2. Machine Learning Engineer
ML Engineers sit deeper in the modeling stack. They build, train, optimize, deploy and maintain machine-learning models.
Programming: Python first, plus SQL; C++ can be valuable for performance-heavy work.
Machine learning: supervised/unsupervised learning, feature engineering, model selection, regression, classification, clustering, recommendation systems, experimentation and model evaluation.
Deep learning: PyTorch, TensorFlow, neural networks, transformers, NLP, computer vision.
Production: Docker, Kubernetes, APIs, cloud computing, data pipelines, model serving and MLOps.
A current August 2026 posting analysis found Python in 66.2% of ML Engineer postings, ML in 62.3%, PyTorch 42.2%, deep learning 32%, LLMs 29.1%, and TensorFlow 28.7%.
3. LLM / AI Agent Engineer
This is becoming an important specialization inside AI engineering: engineers who build systems in which an AI model can reason through tasks, call tools/APIs, retrieve information and perform multi-step workflows.
Skills: Python, LLM APIs, LangChain/LangGraph or equivalent orchestration, agent architecture, tool calling, MCP concepts, structured outputs, RAG, vector databases, embeddings, memory/state management, API integration and workflow orchestration.
But one skill deserves special attention:
AI evaluation.
You need to know how to determine whether an agent actually produced the correct result. Current AI-engineer posting analysis shows evaluations and agentic systems becoming major requirements.
This area could fit particularly well with a software QA/testing background because agent systems require extensive testing, evaluation and reliability engineering.
4. MLOps Engineer / AI Platform Engineer
Think of this as DevOps for AI.
The ML engineer develops the model; the MLOps engineer makes sure it can run reliably in production.
Skills: Python, Linux, Docker, Kubernetes, Git, CI/CD, AWS/Azure/GCP, Terraform, APIs, MLflow or similar tooling, model deployment, monitoring, logging, observability, data/model pipelines and model versioning.
Modern MLOps is also increasingly intersecting with LLMs, RAG and AI applications, rather than dealing only with traditional ML models. Recent job-market data shows MLOps commonly appearing alongside Python, CI/CD, ML, AWS and PyTorch.
5. AI Software Engineer
Traditional software engineering is increasingly becoming software engineering + AI.
You might build a SaaS application where AI is one component rather than the entire product.
Skills: Python and/or Java/JavaScript/TypeScript; APIs; databases; backend development; cloud; distributed systems; Git; testing; CI/CD; system design.
Add:
AI: LLM APIs, RAG, agents, embeddings, vector databases, prompt engineering, evaluations and AI security.
A large 2026 job-posting analysis found AI mentioned in 26.6% of software-engineering postings, alongside established skills such as Java, Python, AWS, React and JavaScript.
6. AI Data Engineer
AI systems are only as useful as the data feeding them. AI Data Engineers create the pipelines and infrastructure that move, clean and organize that data.
Skills: Python, SQL, ETL/ELT, data modeling, APIs, Spark, Kafka, Databricks, Snowflake, cloud storage, AWS/Azure/GCP and data pipelines.
Add AI-specific knowledge around vector databases, embeddings, RAG ingestion pipelines, document processing, training datasets and data quality.
In a 2026 posting analysis, Python and SQL each appeared in more than 60% of Data Engineer postings, followed by data engineering, AWS, Databricks and Spark.
7. Computer Vision Engineer
Computer Vision Engineers develop AI systems that understand images and video.
Applications include autonomous vehicles, manufacturing inspection, medical imaging, security, robotics, facial/object recognition and augmented reality.
Skills: Python, C++, PyTorch, TensorFlow, OpenCV, CNNs, transformers/vision transformers, image processing, object detection, segmentation, image classification, video processing, model optimization and GPU/CUDA fundamentals.
Computer vision also appears as a specialization within ML Engineer postings.
8. Robotics AI Engineer
This is an especially interesting engineering field because it combines AI + software + hardware + mechanical/electrical engineering.
Skills: Python, C++, ROS/ROS2, Linux, computer vision, machine learning, reinforcement learning, motion planning, sensor fusion, control systems, embedded systems and simulation.
For hardware-oriented roles, add electronics, actuators, motors, sensors, CAD/mechatronics and firmware.
Robotics is a significant category in current AI engineering hiring data, and frontier AI companies are actively recruiting across ML, firmware, hardware and robotic-system development.
9. AI Evaluation / AI Quality Engineer
This one is especially worth watching if you’re coming from QA/software testing.
Companies need engineers who determine:
“Does this AI system actually work correctly, reliably and safely?”
That creates a bridge between traditional QA and AI engineering.
Traditional QA skills: test planning, test cases, automation, Selenium/Playwright, API testing, regression testing, Python, CI/CD and defect management.
Add:
AI evaluation skills: LLM evaluation, prompt testing, RAG evaluation, hallucination detection, agent testing, adversarial testing, benchmark datasets, rubric-based evaluation, automated eval pipelines, model regression testing and AI observability.
Evaluation appeared in 56% of AI Engineer postings in one July 2026 analysis—an indication that testing AI behavior is becoming part of mainstream AI engineering rather than a separate afterthought.
10. AI Research Engineer
This is the deepest technical path and usually has the highest mathematical requirements.
Skills: Python, C++, PyTorch/JAX, CUDA/GPU programming, linear algebra, calculus, probability, statistics, optimization, deep learning, transformers, distributed training, reinforcement learning, model architectures and research experimentation.
You’ll also need the ability to read and implement research papers.
This path is much more mathematically demanding than application-focused AI engineering.
The skills that overlap almost everywhere
If you don’t want to commit to one specialty yet, build the common foundation first:
Programming: Python → SQL → Git → APIs
Software: Linux → Docker → CI/CD → cloud
AI: Machine Learning → PyTorch → Transformers → LLMs
Generative AI: Prompting → RAG → Embeddings → Vector Databases → Agents
Production: Kubernetes → MLOps → Monitoring/Observability
Quality: AI evaluation → automated testing → model/agent regression testing
Current job data reinforces this combination: Python remains one of the strongest common skills, while LLMs, agents, cloud infrastructure, data pipelines, PyTorch, fine-tuning and distributed systems are prominent across AI engineering openings.
CURRENT MOONNICHE JOBS
