What skills does an AI engineer need?

Junior skills - call a model from real code

Write clean Python and use an LLM SDK such as openai. Build an API endpoint (FastAPI or Flask) that sends a prompt and returns the answer. Ask the model for JSON and parse it safely. Handle timeouts, rate limits and bad responses with retries. Know the basics of tokens, context size and cost. Use Git and environment variables for API keys.

Mid-level skills - RAG and quality

Create embeddings and store them in a vector database such as Chroma or pgvector. Build retrieval-augmented generation (RAG): split documents, search, and add the results to the prompt. Stream tokens to the user. Cache responses in Redis. Measure quality with an evaluation set and tools like Ragas or DeepEval, and add guardrails for unsafe input.

Senior skills - production AI systems

Design model routing (a small model for easy requests, a large one for hard ones), cost and latency budgets, and fallbacks when a provider is down. Run evaluations in CI so prompt changes do not break quality. Build agents and tool use with clear limits. Know when classic machine learning (scikit-learn, PyTorch) is the better choice than an LLM.

How to show these skills to an employer

Many AI portfolios are demo notebooks. Employers want to see AI inside a working service: an API that calls a model and handles errors, a search feature built on embeddings, a parser that turns model output into reliable JSON. Write each project up with the problem, what you built and how you tested it.

Real AI/ML work to put in your portfolio

Related questions