Skills an FDE needs
32 skills in 6 layers, from code to client. Each page says why it matters on a deployment and how to practise it.
Layer 1 · Code
Python
The default language of data, integrations and LLM tooling.
Stage 1Git
Version control for everything you ship to a client.
Stage 1Testing
Automated checks that prove your code still works after a change.
Stage 1TypeScript
Typed JavaScript for client-side tools and dashboards.
Stage 1Debugging
Finding the cause of a fault in code you did not write.
Layer 2 · Data & integration
SQL
The language of the customer's data warehouse and application databases.
Stage 2REST APIs
How systems talk to each other over HTTP.
Stage 2Webhooks
Event callbacks that a partner system sends to you.
Stage 2ETL pipelines
Extract, transform and load data between systems on a schedule.
Stage 2ERP/CRM integration
Connecting to systems such as ERP and CRM platforms.
Layer 3 · Infra
Docker
Packaging code and its dependencies so it runs anywhere.
Stage 3Cloud basics
Compute, storage, networking and identity on a major cloud.
Stage 3CI/CD
Automated build, test and release pipelines.
Stage 3Auth & secrets
How services prove who they are and where credentials live.
Stage 3Observability
Logs, metrics and traces that show how a system behaves.
Layer 4 · Applied AI
LLM APIs
Calling hosted language models from your own code.
Stage 4Prompt engineering
Writing instructions and examples that make a model behave reliably.
Stage 4RAG
Retrieval-augmented generation: answer from the customer's own documents.
Stage 4Vector databases
Stores that find similar items by meaning rather than keywords.
Stage 4AI agents
Systems where a model chooses tools and steps toward a goal.
Stage 4LLM evals
Fixed test sets and scoring that show whether an AI system is good enough.
Stage 4Guardrails
Checks that keep an AI system inside safe, allowed behaviour.
Stage 4Fine-tuning SLMs
Adapting small language models to a narrow task.
Layer 5 · Design
System design
Designing a system inside someone else's constraints.
Stage 5Data modelling
Shaping tables and entities so the data answers real questions.
Stage 5Security & privacy
Protecting data in transit, at rest and in prompts.
Stage 5Cost optimisation
Keeping model, cloud and engineering spend in line with value.
Layer 6 · Client
Discovery
Learning how the customer's work really happens before building anything.
Stage 5Scoping & SOWs
Turning a wish into a bounded piece of work with agreed outcomes.
Stage 5Stakeholders
Working with the people who sponsor, use and block a deployment.
Stage 5Technical writing
Documentation that a customer team can run without you.
Stage 5Demos
Showing working software to non-engineers in a way they can judge.