What a Forward Deployed Engineer is, and how the work looks in practice
A Forward Deployed Engineer (FDE) is a software engineer who works inside the customer’s operation, next to the people who will use the system, and stays until the software is running in production. The work combines development, integration with the systems the customer already has, and direct contact with the business. My title is CTO of Blanco Brand, and most of my work is done this way.
Where the name comes from
“Forward deployed” is a military phrase for people stationed in the field, close to where things happen. In software the term became known through Palantir, which sent engineers to work inside its customers instead of waiting for requirements to arrive at an office. Today AI companies use it too, because their models have to work inside real processes, with messy data and old systems.
The underlying idea is that the person who writes the code sees the problem first-hand. There is no analyst translating between the operation and the code, and that changes what gets built.
What the work involves
The specifics change from customer to customer, but in my work these keep coming back:
- Spending time inside the operation before writing code. I sit with the people who use the system: a clinic receptionist, a lab chemist, the concierge of a residential community. The problem is rarely the one in the requirements.
- Integrating with what is already there: an ERP that speaks SOAP, Excel spreadsheets, WhatsApp, a health-insurance voucher system. New software has to live alongside what the company already pays for.
- Writing production code, the same code the customer will use, instead of a prototype another team rewrites later.
- Owning what happens after launch: automated tests on every change, fast deployments and an on-call rota with alerting.
What it looks like on real projects
At Alex Stewart’s inspection laboratory, molybdenum concentrate goes through intake, batching, moisture measurement and chemical analysis before it gets a certificate. Each step is done by a different person and leaves a figure the next one depends on. If that chain lives in spreadsheets, a transcription error travels all the way to the certificate, which is the document a sale is settled on. The system we built follows the material through one workflow and generates the certificates as PDF, Word and Excel from the recorded data, with no copying and pasting.
At SuiteMedica, the starting point was watching how patients ask for appointments: a WhatsApp message at eleven at night, a call nobody picks up, an Instagram comment. That led to one AI agent for six channels, which puts the appointment in the clinic’s calendar. I go through it in detail in the note on that agent.
At CARS Control the constraint was that the company’s ERP could be neither replaced nor modified. We built on top: a dashboard and an AI module that read from the ERP without touching it. The pattern works well outside that sector, and I explain it in the note on AI over a legacy ERP.
How it differs from a consultant or a product team
A consultant usually delivers a diagnosis or a prototype and leaves the project. A product team builds for many customers at once and sees each operation from some distance. The FDE sits between the two: building for one specific operation, with code that reaches production, and feeding what is learned there back into the shared foundation.
At Blanco Brand that shared foundation exists. The healthcare systems we build, from SuiteMedica to HemoSuite, ICV OS and CCDM, all have NestJS APIs, and the platform I run has 25 deployed projects. That shared foundation is what lets a problem solved at one clinic carry over to the next.
What it takes to work this way
You need to be comfortable with incomplete information, because the written requirement almost always falls short. It helps to be able to explain a technical decision to someone who is not technical, and to accept that the job is finished when the system works in the operation, which is often later than when the last ticket closes.
You also need a technical foundation that lets you change things quickly. If a deployment takes an hour, nobody ships several times in one afternoon with the customer sitting next to them. On the Blanco Brand platform we brought deployments down from one hour to five minutes, at 99.95% uptime, and every change goes through automated tests before it reaches production.
When it fits and when it does not
It fits when the problem depends on systems that already exist and on how specific people work: a clinic with scheduling, clinical records and billing in separate systems, or a lab where each stage is done by someone different. Being inside the operation there saves months of misunderstood requirements.
It fits less when what is needed is a standard product that only has to be configured. Sending an engineer into the operation is expensive in that case and adds little.
If you are curious how this way of working combines with a CTO title, I wrote about it in Hands-on CTO.
If you work this way, or are thinking about organising a team like this, I would like to compare notes. Write to me at hola@bcristi.com.