DeepLearning.AI — Forward Deployed Engineers and the Future of AI Engineering¶
Source: DeepLearning.AI — The Batch by Andrew Ng
Andrew Ng examines the rapid rise of Forward Deployed Engineers (FDEs) — software engineers embedded directly at client organizations to customize and deploy AI solutions.
The FDE Surge¶
The FDE role, pioneered by Palantir over 20 years ago, is having a moment. Current signals:
- OpenAI launched a dedicated "deployment company" focused on enterprise customization
- Anthropic partnered with Blackstone, Hellman & Friedman, and Goldman Sachs to embed FDEs in financial services
The promise is clear: a model alone isn't a solution. You need engineers who understand both the AI stack and the client's domain to bridge the gap.
The Inevitable Shift¶
But Ng argues this is a transitional phase. As AI engineering matures, companies will increasingly want their own AI engineers rather than borrowed ones from vendors. The FDE model works for early adoption, but long-term, every enterprise will need internal AI engineering capability.
Future Specializations¶
Ng identifies the emerging sub-disciplines of AI engineering:
- LLMOps — operating and maintaining LLM systems in production
- Evals — evaluation frameworks for measuring AI system performance
- AI Data — data pipelines, curation, and synthetic data generation
- Harness Engineers — building the infrastructure (the "harness") that connects models to tools and data
The Bigger Picture¶
AI is creating jobs, not destroying them — but the jobs are new and different. The FDE role is just the first of many AI engineering specializations that will emerge as the field industrializes. The key insight: AI engineering is becoming a distinct discipline, separate from both traditional software engineering and ML research.