Original author: @AndrewYNg
Compile: Peggy
Editor’s Note: As companies like OpenAI and Anthropic begin forming teams of AI Forward Deployed Engineers (FDEs), an older role originating from Palantir is gaining renewed popularity in Silicon Valley. The core value of an FDE lies in going on-site to customers and transforming general-purpose large models into agent workflows tailored to specific business processes.
But what this article truly discusses is not just the new profession of FDE, but how job structures are being reshaped in the AI era. The author argues that, compared to the small number of FDEs deployed at client sites to support specific vendor product implementations, the greater future demand will be for in-house AI Engineers. These professionals need to understand prompts, agent frameworks, and evaluation systems, and be proficient in using AI programming tools like Claude Code and Codex to truly embed AI capabilities into software and business systems.
This also means that AI’s impact on the job market isn’t simply about “replacement.” It’s more likely to first create a new wave of generalist roles, then evolve further—just as software engineering once split into frontend, backend, mobile, and DevOps—into more specialized positions like LLMOps, Evals Engineer, and AI Data Engineer. What will truly be in short supply are individuals who understand both engineering implementation and business contexts.
The following is the original text:
Recently, a highly discussed new role has emerged in Silicon Valley: the AI Forward Deployed Engineer (FDE). These engineers are embedded within client organizations to help customize solutions, such as building and optimizing Agent workflows tailored to the client’s specific needs. Since OpenAI and Anthropic began forming new teams to deploy FDEs directly into client organizations, I’ve heard many people rediscover this career path.
AI workloads are driving the rise of FDE roles, illustrating how AI is creating new jobs. This also shows that the narrative of an impending job market collapse—the so-called “jobpocalypse”—is unfounded; there will still be abundant AI-related and non-AI-related roles in the future. However, as explained below, I believe the number of AI engineer positions will far exceed that of FDEs.
The role of FDE was pioneered by Palantir about two decades ago. At that time, Palantir would send engineers on-site to government agencies, working in secure, air-gapped environments. In addition to strong technical skills, FDEs needed communication abilities and sometimes business judgment—for example, engaging with clients to understand their needs, setting project priorities, explaining complex technologies, and providing respectful yet firm feedback when clients made unrealistic requests. Today, FDEs are once again gaining attention, primarily because truly embedding a ready-made large language model into enterprise operations and transforming it into a customized agent workflow tailored to specific business needs requires substantial hands-on implementation work.
However, I believe the scale of AI engineer roles will be much larger. A company might accept a small number of FDEs for internal collaboration, but most companies will want more of their own employees involved in project development. Take my organization as an example: while we do hire FDEs, we hire far more AI engineers. Additionally, a common concern among clients is that it’s difficult to find truly “vendor-neutral” FDEs. After all, the core function of an FDE is to deeply integrate a specific vendor’s product into enterprise systems. At this stage, it’s hard to predict which AI service will be the best choice a year from now, making “optionality” crucial—the ability for enterprises to choose the most suitable vendor in the future. In contrast, having FDEs tightly couple a company’s business processes with a single vendor significantly reduces this optionality.
Currently, I’m seeing a rapid increase in market demand for AI engineers—professionals who can build applications using AI software components such as LLM prompts, agent frameworks, and evaluation systems, while also efficiently leveraging AI programming agents like Claude Code, Codex, Antigravity CLI, and OpenCode. As the role of AI engineer matures, I expect it to further split into more specialized positions, much like how the general term “software engineer” diversified decades ago into frontend, backend, mobile, data engineering, DevOps, and other domains.
What specialized AI engineering roles will emerge in the future? I’m not certain yet. There might be AI FDEs, LLMOps engineers, evaluation engineers, AI data engineers, Harness engineers, and some new roles we haven’t even named yet. But at least for now, many generalist AI engineers are already creating tremendous value. Skilled AI engineers are in high demand. As this field continues to mature over the next decade, I also expect greater specialization within AI engineering, leading to the creation of even more new job opportunities.
Click to learn about the open positions at BlockBeats
Welcome to the official BlockBeats community:
Telegram subscription group: https://t.me/theblockbeats
Telegram group: https://t.me/BlockBeats_App
Official Twitter account: https://twitter.com/BlockBeatsAsia