What an L7 engineer actually does
An L7 (senior staff) engineer owns a bet, not a system. The work is choosing which problem is worth solving across an organisation and being right often enough that the reasoning stops being questioned. Technical depth is assumed at this level; what distinguishes it is judgement under genuine ambiguity.
The clearest way to see the ladder is by what is being judged. At L5 the design is judged on whether it works. At L6 it is judged on what it costs the teams around you and on what it forecloses. At L7 the problem selection itself is what is judged - you are trusted to say which of five plausible directions is the one worth two years, and that trust is the level.
This is why L7 is hard to study for directly. You cannot revise your way into judgement. What you can do is remove the specific gaps that stop you being credible on a given surface - and right now the most common one, for engineers who are otherwise ready, is being unable to reason about what a model can and cannot do.
The material below is the honest version of that gap: enough mathematics to read the field, the mechanics of how models learn, transformers, retrieval, and the operational side of running any of it in production.
What to actually study
Real modules from the curriculum, in the order they make sense. Every video each one points at is free.
- Inference EngineeringThe Generation LoopUnderstand how an LLM produces text token by token --- from the forward pass and logits through sampling and stopping to the serving stack that puts it all in production.11 lessons
- Inference EngineeringThe KV CacheDesign and measure the single most important inference optimisation. From why caching exists through the arithmetic, the memory wall, attention variants built to shrink it, fragmentation and paged allocation, prefix caching and radix trees, to hit rate as the production metric that drives both cost and latency.11 lessons
- DeploymentThe Language of SystemsThe five behaviours every production system exhibits, stated as maths before they are stated as advice.6 lessons
- Inference EngineeringThe Hardware FloorMeasure and model GPU performance from first principles --- throughput vs latency, FLOPs and bytes, the memory hierarchy, arithmetic intensity, the roofline model, and why decode is memory bound. By the end you can read a profiler trace and name exactly which regime your bottleneck lands in.12 lessons
- Everything DataData Versioning & ML ReproducibilityMake datasets, features, experiments, and model outputs reproducible across time, teams, and environments.7 lessons
Common questions
- What is an L7 engineer at Google?
- L7 is senior staff software engineer. It sits above L6 (staff) and is an organisation-wide scope: the expectation is that you choose and justify technical direction across multiple teams rather than design within a direction someone else set.
- Does L7 mean managing people?
- No. L7 is a rung on the individual-contributor ladder and runs parallel to management. The influence is technical - what gets built and why - rather than headcount.
- Can a course make me an L7 engineer?
- No course promotes anyone. Levels are awarded by employers against their own ladder, on evidence collected over months. What material like this can do is close a specific knowledge gap that is currently in the way.
No course promotes anyone. Levels are awarded by employers against their own ladder, on evidence collected over months. What material like this can do is close a specific knowledge gap that is currently in the way.
Every level