Careers in artificial intelligence: roles, skills, and market opportunities

Working with AI isn’t just about building models from scratch. Behind tools like ChatGPT and Claude are professionals who train systems, test responses, identify security flaws, and develop agents capable of performing tasks with varying levels of autonomy.

These activities require a diverse set of skills, ranging from programming and statistics to critical analysis, business knowledge, and digital security. Job postings from companies like OpenAI and Anthropic illustrate how the industry combines different specialties to build and apply this technology.

1. Model training: teaching AI to respond better

Model training is one of the most technical areas of AI. Here, researchers and engineers manipulate data, experiments, and mathematical methods to create systems that recognize patterns, follow instructions, and perform tasks.

A key step is post-training, which improves an existing model. This process can include examples of desired responses, human or automated evaluations, and adjustments to enhance the system’s behavior.

For instance, OpenAI offers positions related to agent training and developing techniques to improve their capabilities. These roles typically require experience in machine learning, software engineering, statistics, or related fields.

2. AI evaluation: testing responses and finding flaws

Not every convincing response from an AI system is correct. That’s why another professional area focuses on measuring the quality, consistency, and safety of the results.

These professionals create tests, compare responses against established criteria, and investigate situations where the system makes errors. This process is known as model evaluation, or AI evaluation.

A related activity is red teaming, which involves deliberately testing a system’s limits to uncover vulnerabilities. In a corporate setting, the test might check whether an assistant discloses confidential information or follows malicious instructions.

The work can involve researchers, security engineers, and specialists in the domain where the AI will be used. The Organisation for Economic Co-operation and Development (OECD) describes adversarial evaluation as one of the practices developers use to identify risks in systems.

3. Agent engineering: creating systems that perform tasks

AI agents are systems capable of planning steps, using tools, and executing sequences of actions to achieve a goal. Unlike a chatbot that only answers questions, an agent can, for example, consult sources, organize information, and interact with other software.

Developing these systems involves professionals who design execution flows, connect models to external tools, and verify that tasks are performed reliably. Depending on the application, it’s also necessary to establish access limits and human supervision mechanisms.

OpenAI has openings related to agent engineering and applying AI in companies’ products and processes. In São Paulo, the company also advertises applied engineering roles that combine technical development, model evaluation, and implementing solutions for clients.

4. What skills do these careers require?

Requirements vary by role. Positions focused on research and model training typically demand advanced technical knowledge. Applied engineering roles, on the other hand, combine programming, system integration, and an understanding of user needs.

In evaluation, analytical reasoning, test design, and knowledge of the usage context help identify errors that isolated metrics might miss. In security, it’s crucial to understand how systems can fail or be manipulated.

There are also opportunities in non-technical areas. Anthropic, the company behind Claude, says it’s looking for professionals in public policy, operations, and business, in addition to researchers and engineers. The company highlights the diversity of experience among its technical staff, though requirements vary by position.

How to develop AI skills for work

Understanding AI fundamentals also helps professionals in other fields grasp how the technology can be applied in their work. Understanding model limitations, formulating clear instructions, and evaluating results are valuable skills for those in communication, marketing, operations, or management.

For those looking to get started, Saint Paul’s Pre-MBA in AI offers four practical online classes for R$ 37 as an introduction to the topic.