Home/Guides/Data & AI
Data & AI

Data & AI: SFIA levels and developer seniority

Fennec team · 1 Jun 2026 · 4 min read

Data and AI work moves fast enough that the specific tools shift every few months, but the underlying levels of responsibility don't. Someone running a pre-built model on their own data and someone deploying and monitoring ML systems in production are at genuinely different levels, whatever the job title says.

SFIA levels here track a move from experimenting with existing models, to building ML pipelines end to end, to setting AI strategy and governance for an organisation.

Machine Learning
TensorFlow / PyTorch
Data Science
Pandas

1

Level 1: Follow

“I understand ML concepts and have experimented with pre-built models or tutorials. I know what training data means.”

What it looks like

Evidence here is exploratory by design: notebook or tutorial work using a pre-built model, a completed ML fundamentals course, or notes explaining what training data actually means in your own words rather than borrowed ones.

What moves you forward

Run a pre-built model on your own dataset rather than only the tutorial's, that's where the real learning happens. Learn what training versus inference actually means hands-on, and work through one applied ML course that goes beyond the theory you've already covered.

2

Level 2: Assist

“I prepare datasets, run existing training pipelines, and evaluate model outputs using provided metrics.”

What it looks like

The evidence shifts toward hands-on preparation: a dataset you prepared end to end, cleaning and splitting included, an existing training pipeline you ran and evaluated, or model outputs you assessed against provided metrics without needing someone to interpret them for you.

What moves you forward

Prepare a dataset end to end, cleaning, splitting, and validation, as a deliberate exercise. Run an existing pipeline and explain the metrics it produces in your own words, and learn to read a confusion matrix, or the equivalent for your problem type, well enough to spot when something's off.

3

Level 3: Apply

“I build ML pipelines end-to-end, from data prep through training to production deployment with monitoring. I select appropriate algorithms, design experiments properly, and document results reproducibly.”

What it looks like

Here the evidence is a full pipeline: an ML pipeline you built end to end, from data prep to a monitored production deployment, an experiment you designed and documented well enough for someone else to reproduce, or a model choice you made and could actually justify.

What moves you forward

Take one model from notebook to a monitored production deployment, that gap is usually where the real learning lives. Document an experiment so someone else could reproduce it, and pick the right algorithm for a problem and write down why, since that reasoning is what people will actually ask you about later.

4

Level 4: Enable

“I lead ML projects from problem definition to production. I translate messy business questions into well-defined ML problems, implement MLOps practices, and help data scientists on my team do better work.”

What it looks like

By now the evidence is leadership of the problem, not just the model: an ML project you led from a business question to production, an MLOps practice, versioning, monitoring, or retraining, that you implemented, or a data scientist you helped do meaningfully better work.

What moves you forward

Translate a vague business question into a well-defined ML problem yourself rather than waiting for someone else to frame it. Implement one MLOps practice your team was missing, and review another engineer's experiment design, since spotting flaws in someone else's approach is a different skill from avoiding them in your own.

5

Level 5: Ensure & Advise

“I define AI/ML strategy for the organisation. I lead large-scale AI programmes and govern responsible AI practices and ethics.”

What it looks like

The evidence spans the organisation: an AI/ML strategy you defined, a large programme you led across multiple teams, or a responsible-AI review process that exists because you put it in place.

What moves you forward

Write an AI/ML strategy document rather than leaving the direction implicit, and lead a programme spanning multiple teams. Setting a responsible-AI review process is increasingly part of this level, not a side project, as the stakes of getting AI wrong keep rising.

6

Level 6: Initiate & Influence

“I set AI strategy and responsible AI governance at an organisational or industry level.”

What it looks like

At the top, the evidence is strategic and often public: organisation or industry-level AI strategy and governance you set, external recognition through research, talks, or published work, or a responsible-AI standard that shaped decisions well beyond your own team.

What moves you forward

From here, publish your AI strategy thinking externally, set AI governance at the organisation level, and expect to represent AI strategy at the leadership table as much as you touch a model directly.

Go deeper

Hugging Face blog

Practical writing on models, training, and deployment.

Papers With Code

Research papers paired with their reference implementations.

Lex Fridman Podcast

Long-form conversations with AI researchers and practitioners.

Knowing where you sit is one thing, proving it later is another. Fennec lets you log data & ai evidence as you go, a shipped feature, a decision, a review, tagged to the level it demonstrates, so the case for your next step is already made when you need it.

fennec
Developer growth, tracked and proven.