Back to journal

Career · · 1 min read

Five kinds of work around an AI product

Read the responsibilities behind an AI job title before choosing a learning path.

Five kinds of work around an AI product

Expanded from my Telegram note. The English edition is a translation and editorial adaptation.

While browsing AI job descriptions, I noticed how much work happens around a model: understanding a customer, connecting systems, defining tools, measuring quality and testing boundaries.

The five roles I shared are a snapshot of that browsing. Titles and vacancies change, so the more durable question is what responsibility each role carries.

From customer problem to working system

A forward deployed engineer works close to the customer and helps turn a practical need into a deployed solution. An applied AI architect considers how that solution fits data and existing systems.

An agent engineer focuses on the agent’s tools, control flow and error handling. The boundaries between these jobs can overlap substantially across organizations.

Quality is a job too

Product evaluations work asks whether a change actually improves the system and which earlier successes it breaks. Red-team work probes misuse paths and failure boundaries.

These responsibilities need different combinations of engineering, communication and analytical thinking. A title alone does not tell you which combination a team expects.

Choose a small project

If you are early in your career, build one small application and keep a set of evaluation examples. Document its tools, failure handling and limits. Then compare the work you enjoyed with real job descriptions. That gives your next learning decision more substance than collecting framework names.

Ali Rashedi