Top AI skills listed

A viral career thread is circulating a simple claim: companies are paying for three kinds of artificial intelligence work now — writing better instructions, adapting models, and wiring them to outside knowledge. (x.com) The first skill is prompt engineering, which means writing and testing instructions so a model gives the format, tone, and facts a team needs. OpenAI says prompting is the process of providing input to a model, and that output quality often depends on how well the model is prompted. (developers.openai.com) The thread put prompt-engineering pay at $80,000 to $150,000 a year. Current job ads show that range is plausible but uneven: DeVry listed an Applied AI Prompt Engineer role at $100,000 to $115,000, while McKesson posted an AI Engineer role that includes prompt-engineering work at $104,600 to $174,400. (indeed.com 1) (indeed.com 2) The second skill is fine-tuning, which is like giving a general model extra training on a smaller, targeted dataset so it performs better on one domain or task. Hugging Face says fine-tuning continues training a large pretrained model on a smaller dataset specific to a task or domain. (huggingface.co) The thread put artificial intelligence and machine learning fine-tuning roles at $120,000 to $200,000 a year. Recent postings overlap that band: Tata Consultancy Services listed an AI and machine learning engineer in Atlanta at $140,000 to $160,000, and Philips listed a data and artificial intelligence scientist role in Cambridge at $129,000 to $205,000. (indeed.com 1) (indeed.com 2) The third skill is retrieval-augmented generation, usually shortened to RAG, which means a model looks up relevant documents before it answers instead of relying only on what was in its training data. OpenAI says retrieval uses semantic search to surface similar results from a company’s own data and is especially useful when combined with models to synthesize responses. (developers.openai.com) That work is showing up directly in job titles. Microsoft posted a Member of Technical Staff role focused on retrieval-augmented generation at $139,900 to $331,200, while Elsevier listed a Senior Software Engineer for retrieval-augmented generation at $95,300 to $158,800. (indeed.com) The thread’s advice to practice on ChatGPT and Hugging Face lines up with the tools employers already expect people to know. OpenAI publishes prompt-engineering guides for its models, and Hugging Face runs a free Learn hub with courses on large language models, agents, and an open-source cookbook. (developers.openai.com) (huggingface.co) Hugging Face also publishes step-by-step notebooks for advanced retrieval-augmented generation, including a build that answers questions over Hugging Face documentation with LangChain. That makes the learning path concrete: prompt a model, fine-tune a model, then connect a model to a document store and test the answers. (huggingface.co) The jobs market around retrieval alone is broad, even if many listings use adjacent titles instead of “RAG developer.” Indeed returned more than 4,000 jobs for “RAG retrieval augmented generation” and more than 25,000 for the full phrase “retrieval augmented generation,” though those searches also capture some unrelated uses of the word retrieval. (indeed.com 1) (indeed.com 2) The thread landed because it turned a blurry “learn artificial intelligence” pitch into three specific job buckets with salary bands and free practice tools. The harder part is not finding resources in April 2026; it is building projects that show you can make a model answer better, cheaper, or with fewer mistakes. (x.com) (developers.openai.com)