Neeraj Sujan

Forward Deployed Engineering · Production AI · Inference

The deployment gap
is the real job.

AI rarely fails because the model is bad. It fails in the last mile — messy data, unclear ownership, no trust, no rollback. This is about the engineer who closes that gap: forward deployed, customer-facing, building AI that ships into reality.

What this covers

Forward Deployed Engineering

The FDE role decoded — ambiguity tolerance, customer discovery, owning outcomes end-to-end, and the interview system to get the job.

Production AI Systems

The prototype-to-production gap. What breaks when your demo hits a real customer environment, and how to build AI that survives contact with reality.

Inference & Model Serving

Latency, throughput, batching, quantization, and the engineering decisions that make AI fast and cost-effective at scale.

Evals & Reliability

Building evaluation systems that actually catch failure. Observability, trust mechanisms, and the feedback loops that make AI systems trustworthy.

Customer-Facing Engineering

Discovery frameworks, trade-off communication, stakeholder navigation, and the skills that make a technical person effective in a customer room.

Recent writing

All posts →
Your AI Worked in the Demo. Here's Why It Dies in Production.Great Engineers Freeze in FDE Interviews. Here's the Fix.Why the Best Coder Is No Longer the Most Valuable Engineer

Series

Newsletter

Writing on FDE, production AI, and the engineering discipline of deploying AI into reality. No schedule — only when there's something worth saying.

No spam. Unsubscribe anytime.