Last update: July 2026. All opinions are my own.

GenAI Engineering — Interview Prep · Part 2/15

The classic bad answer to "why do you want to work here?" is a warmed-over version of the company's About page. This is my honest version.

Why GFT specifically (not just "any AI role")

1. AI-Centric is a real position, not a slogan.

GFT's public framing is "pionera en transformación digital · soluciones de negocio centradas en la inteligencia artificial". The word centradas is doing the work there. Most consultancies bolt AI onto existing engagements. GFT positions AI as the architecture the engagement is built around, which is a very different thing operationally — it means the RAG pipeline is not an add-on to a data warehouse project, it is the project.

That framing matters for me because it means the conversations I want to have — about retrieval quality, prompt injection defense, agent evaluation, cost per query — are first-class questions in the room, not something I have to fight to get on the agenda.

2. The Center of Excellence (AI.DA) model.

Consultancies that don't have a CoE tend to have every engineer solving the same problems in parallel across projects. Consultancies with a CoE have a shared playbook, shared tooling, shared learnings — which means the multi-agent pattern someone shipped in Q1 becomes the starting template for the next three engagements.

For me, that's the difference between "learn everything from scratch on every project" and "learn from senior architects while shipping". The second is where I want to be right now.

3. Banking + insurance + industrial is where GenAI has to work under real constraints.

The interesting engineering problems in GenAI are not "can we make a chatbot answer trivia". They are: can we make a RAG system that satisfies a bank's audit requirements, doesn't hallucinate on regulatory questions, and stays under a per-query cost budget when the compliance team asks it 400 questions a day?

That is the exact class of problem I want to work on. It's real. It's constrained. And it means every technical choice — chunking strategy, guardrail design, model routing — has a business consequence I can point at. That's the environment where I learn fastest.

4. Python + Java + Kafka + microservices is the stack I want to grow into.

The JD is honest that this is a polyglot engineering environment. Python is my primary language and I'm strong there. Java + Kafka + microservices is the growth zone — and it's the growth zone I want, because the future of GenAI engineering is not notebooks, it's event-driven distributed systems with LLMs as one component. GFT sits at that intersection.

5. Multi-agent patterns as a first-class skill area.

Most JDs list "GenAI experience" and mean "you've used ChatGPT". GFT's JD lists multi-agent patterns (skills + subagents), reasoning loops, tool design, error handling in tools — the modern agent-engineering vocabulary. That tells me the team is already past the demo phase and thinking about how to make multi-agent systems robust in production. That is exactly the frontier I want to be working on.

6. The stated engineering culture.

The "competencias personales" section reads: trabajo en equipo, favoreciendo dinámicas que permitan el flujo de conocimiento, con interés tanto en aprender como en enseñar. That's a real signal. It means the culture explicitly rewards teaching and learning as part of the job, not something you do on the side. My blog is essentially teaching-in-public. That aligns.

7. Madrid / Barcelona / Valencia — real cities with real teams.

The location list (Madrid, Alicante, Lleida, Sant Cugat, Valencia, Zaragoza) tells me GFT has substantial presence across Spain, not one HQ with satellite offices. That means whichever office I'd land in has a real team locally, not a screen-projected remote hub.

8. English at B2/C1 is valorable, not required.

I appreciate that they say "valorable" — the primary work happens in Spanish. Working in a Spanish-first engineering environment on international projects is the exact professional level I want to be at right now. I'm bilingual enough to add value on international loops without pretending my English is native.

What I'm bringing that fits the CoE

  • A visible teaching + writing habit that translates directly into knowledge-sharing inside a CoE.
  • A ML/NLP foundation strong enough that I can hold a technical conversation about why the modeling choices are what they are, not just what the frameworks default to.
  • A learning velocity that closes gaps in weeks, not months.
  • The interview-prep series itself as evidence of #1 and #3 combined — I identified a gap (production GenAI), studied it structurally, and published a resource that other people can now use.

What I want from GFT that I can't get elsewhere

  • Senior AI architects to learn LLMOps from, in production.
  • Real client problems where the constraints (compliance, cost, SLA) are not made up.
  • A CoE structure that means my learning becomes team learning.
  • A polyglot stack that pushes me past notebook-only Python.
  • A culture where teaching is part of the job description.

What I'd say if they ask "why GFT?" in the room

"Two things specifically. First, the AI.DA Center of Excellence model — I'm at the point in my career where I want to ship production GenAI systems, and the CoE format gives me both the senior architects to learn from and the shared playbook to build against. Second, the client focus — banking, insurance, industrial — is where GenAI has to work under real regulatory and cost constraints. That's the environment where I learn fastest, because every design choice has a consequence you can point at. And honestly, the JD reads like it was written by a team that has thought carefully about what modern GenAI engineering actually looks like — multi-agent patterns, reasoning loops, guardrails, LLMOps as first-class concerns. That's a team I want to learn from."

What's next in this series