Skip to content
SciTechBiz Institute
I want to join
Blog · Forward Deployed Engineer

From building with Claude or ChatGPT to Forward Deployed Engineer: what changes

Moving from building with Claude or ChatGPT to Forward Deployed Engineer changes where you work and how your work is measured. You can already put a demo together in an afternoon. What is added is building inside a company, with its data and its team: measuring with real cases, passing security and leaving the system running without you.

Updated: 5 November 2026 · Deibson Silva · Forward Deployed AI Engineer and Founder

Where you start

What can you already do if you build with Claude or ChatGPT?

If you build with Claude or ChatGPT, you describe what you want in words, read what the model writes and correct it until it works. People call it vibe coding, and it is useful: in an afternoon you have an assistant, a dashboard or an automation that used to take weeks.

That skill counts. Splitting a request into pieces, spotting when the model is wrong and asking for another version is part of the craft of anyone who builds with AI.

The limit shows when what you made has to serve a company. The demo works on your screen, with the examples you chose. In the client's office, documents arrive as crooked photos, suppliers send the same thing twice and there are exceptions known only to the person who has done that job for years.

One scene keeps coming back. Someone puts an AI assistant together in an afternoon, the demo goes down well and nobody knows how to measure whether it gets things right. Nobody has written down what it may do alone and what it may not. Three months later it is still there and hardly anyone uses it.

Comparison

What changes between building on your own and building inside a company?

The tooling can be the same. Almost everything else changes.

Building on your ownForward Deployed Engineer
Where you workOn your computer, with your examplesInside the client company, alongside the team that will use the system
Whose judgement it isYours: you decide what is rightThe in-house expert's: you write it as rules that person reviews and approves
How you know it worksYou try it by hand and it looks rightYou measure it with a suite of 20 to 50 tasks drawn from real failures
What happens with securityIt is looked at last, if at allYou attack the agent on purpose before release and document its EU AI Act sheet
Who uses itYouThe client's team, in their daily work
When you are doneWhen the demo worksWhen the company runs it alone, after the handover
The method

Which stages does the method add to what you already do?

Prototyping is one of the seven stages of the Legathum AI-First Hybridisation Method v3, and it is the one you already know. The other six are what changes:

  1. 1Qualify: before you open the editor, you decide with the company whether the case deserves AI. A vague «we want AI» becomes a problem with an owner, a metric and a baseline, meaning how things stand today.
  2. 2Discover in the field: you sit with the person who does the work and look at real cases. That is where the things nobody mentions in a meeting turn up.
  3. 3Capture the criterion: you write, in plain language, the rules the in-house expert decides by, each with its source, and that person approves them.
  4. 4Prototype in a day: you build a first version with the team and the case goes ahead or stops. Stopping is a valid outcome too.
  5. 5Build with evals: you measure the agent with real cases before anyone uses it, and attack it to see where it breaks.
  6. 6Deploy and adopt: you put it into production in phases and check that the team uses it day to day.
  7. 7Hand over: you leave documentation, tests and named owners so the company runs without you.

Six tests signed by the client mark progress: Loop, Bearer, Mirror, Security, Adoption and Handover.

What is new

What do you have to learn that the model will not do for you?

The coding agent writes. Deciding what gets built and whether it is right stays with you.

Give it a way to check itself

A coding agent works better when it can verify its own work. The test is written first and acts as the specification: it fails at the start and passes once the change is right.

Read before you write

The cycle is explore, plan, code and verify. The plan is reviewed before a single line of code exists, and changes are saved in small steps that can be undone.

Review what you did not write

Nothing reaches production without a review by another person, or by an agent other than the author. Reading someone else's code to find faults is learnt by doing.

Look after secrets and data

A credential lives in the environment, never in the code. Personal data is masked before it is shown, and test data is always synthetic.

Profile

Do you need to know how to code?

There is no entry test. The Forward Deployed Engineer Licence is for people who already build with Claude or ChatGPT, whether or not they can code, and also for engineers, data analysts and heads of operations or product.

You build with Claude Code as a copilot. Python, Node and git are installed with you in the first week.

What the programme asks for is logic and review: breaking a problem down, reading carefully what the agent proposes and checking it. Easy it is not. It ends with a four-week project at a real company and a 30-minute oral defence before Deibson Silva.

The programme

How do you get from one to the other?

Through the Forward Deployed Engineer Certification programme of the SciTechBiz Institute, which you take inside the FDE Licence. It lasts 12 weeks: eight of modules and labs, and four of project.

The programme has 13 modules, 80 lessons and 12 labs. Each lab checks itself and uses synthetic data. One is an application built with Claude Code and tests, one is a suite of evals and regression, one is a simulated attack and one is a handover kit.

The EU AI Act runs through every stage. Article 50 has applied since 2 August 2026: people talking to an AI system must be told it is an AI, unless that is obvious from the context. That is why each agent carries its own sheet, with its risk, its tests and its human oversight.

People who pass are awarded the title of AI-First Hybridisation Architect, with a digital certificate and a verification code.

The FDE Licence is an annual licence. From the day you buy, it gives you the Organizational Consultant's rights on the Community's platforms, and the programme and its assessment come with it.

Good to know

Honest limits

  • The certification attests a competence, never a job or an income.
  • Prototype in a day is a stage of the method. No FDE promises the client a timescale for results.
  • Today the programme is taught in Spanish: 27 Spanish videos are published at the Academy, and the English videos arrive in November 2026.
  • To deliver to client companies you need a valid FDE Certification and a signed Framework Contract and data protection agreements. The FDE annex to the contracts is being prepared.
  • The labs use synthetic data only. In the real project, the company's data never leaves its systems.
  • The SciTechBiz Institute is not a university: its certifications are awarded by the Institute itself and are not recognised by any education authority.
FAQ

What people ask most

01

What is vibe coding?

It means building software by describing what you want to an AI model, such as Claude or ChatGPT, and correcting what comes back. It gets you a prototype quickly. For that prototype to work in a company, it still needs tests with real cases, security and a handover.

02

Do I need to know how to code to become a Forward Deployed Engineer?

There is no entry test. You build with Claude Code as a copilot, and Python, Node and git are installed with you in the first week. What is asked of you is logic and review: you decide what gets built and whether it is right.

03

What is the difference between a demo and a system in production?

A demo gets it right with the examples its author chose. A system in production has been measured with real cases, has passed a security test, is used by the team in their daily work and has named owners inside the company.

04

How long does the programme take?

12 weeks: eight of modules and labs, and four of project at a real company, with its written consent. It ends with a 30-minute oral defence before Deibson Silva.

05

What exactly do you buy with the FDE Licence?

An annual licence. It gives you the Organizational Consultant's rights on the Community's platforms from the day you buy, and the programme and assessment of the FDE Certification come with it. The certificate is issued when you pass the assessment.

Portrait of Deibson Silva
Deibson Silva · Forward Deployed AI Engineer and Founder

Founder of Legathum and iPersonality and Forward Deployed AI Engineer. He leads the SciTechBiz Community from Madrid and examines the FDE Certification defence.

Your licence

Start from your dashboard

On the licence page you request access; requesting access costs nothing. You then sign the contract and pay the annual licence from your dashboard.

iPersonality Global Institute, S.L. · NIF B27580042 · Madrid · Privacy policy

Prices in euros, exclusive of VAT. In Spain, Spanish VAT is added. In the rest of the EU, VAT is calculated at payment according to your country and VAT number. No Spanish VAT is charged to UK customers.