In a newly released research project examining nearly 50,000 job postings, Andela Head of Research Corey Hymel analyzes how artificial intelligence is transforming developer roles and technical skill requirements.
Mapping the Rise of LLM Application Engineers
Enterprise hiring data reveals distinct and specialized skill bundles forming as net new roles or as underlying skills within historical roles change. Hymel and his research team tracked these emerging definitions. Job titles have historically lagged behind actual workflow changes, but current hiring demand is driving specialization.
One primary structural shift highlighted in the research is the emergence of the LLM application engineer. This role is for someone versed in building products on top of LLMs, whether it’s OpenAI APIs, open weight models, or Claude code. They know how to plug into the APIs, but they can also construct the harnesses and are familiar with evals, governance, and guardrails—a specific skill set that differs from a classical backend engineer.
Balancing Costs With FinOps Reliability
Another major development is the rise of the FinOps reliability engineer. As enterprises build their AI-native software development lifecycles, managing consumption-based spending has become a priority. These specialized engineers are able to think about project development and factor in the cost of end-user consumption, cloud consumption, and developer tooling.
Agentic Workflows and Docs-as-Code
The research also maps out the docs-as-code engineer.
Confronting Enterprise Talent Debt
The Andela research introduces the concept of talent debt to describe the rapid accumulation of skill gaps within enterprise engineering teams. Similar to technical debt, talent debt grows as the half-life of relevant technical skills shrinks substantially under the weight of accelerating technology change.
Engineering organizations facing this deficit currently must choose between two options. They can reskill existing personnel—preserving deep domain and system knowledge—or they can recruit externally, which is faster but risks ramp-up delays and the risk of putting the wrong person in the wrong seat.
