apprenti-dev/software-engineering-base

The first official apprenti.dev path — 54 tasks across four academic years plus summer practice, covering software engineering from first principles through independent, mentor-reviewed work.

Year 1 (both terms plus summer) and Year 2 Term 1 so far — 19 tasks moving from analytics and inference through first models to LLM and retrieval work, each evaluated on held-out cases, not vibes. Later years aren't authored yet.

For working developers. Treat AI assistance as disclosed, verified professional practice — agentic workflows, review of generated changes, and knowing when not to reach for a model.

For junior to mid-level developers specializing into services. Design, build, test, operate, and defend one coherent backend — HTTP APIs, data, messaging, containers, and CI.

For mid to senior engineers. Non-functional requirements, defensible ADRs, capacity and SLO models, and defending a decision when a constraint changes.

For new graduates and juniors who can already code. Professional Git, reading unfamiliar code, testing what exists, methodical debugging, and writing for a reviewer.

Build a real pipeline, ship through it, observe what happens, and roll it back. Containers, CI, infrastructure as code, and production signal — locally first.

For mid and senior backend engineers. Replicate, partition, and break systems until consistency and failure trade-offs are something you can reproduce, not folklore.

The engineering of data movement — batch and stream pipelines, contracts, orchestration, idempotent reprocessing, schema evolution — distinct from the analytics and modelling path above.

Defensive AppSec for software engineers: threat modelling, authentication, secure coding, secrets, and verified fixes. Find and fix issues in work you own — not a CTF.

Map tests to real risk, measure quality directly, and defend the gates you add to a pipeline — contract tests, mutation testing, load tests, CI quality signals.

For seniors and tech leads. Mentoring, delegation, cross-team influence, and technical decisions at team scope — judged on inspectable briefs, reviews, and communications.

Measure first. Profile, benchmark, cache with evidence, load-test, and reason about capacity from numbers you produced — not guessed.

Instrument, measure, alert, respond, and learn without blame. Logs, metrics, traces, SLOs, and incident response as evidence, not tooling tourism.

Real contribution and maintainership. Evidence is an externally verifiable pull request, issue, or review — licensing, collaboration, and judgment under your own name.

Resource-constrained and edge software: determinism, constrained memory, industrial protocols, and update safety. Host-side C, QEMU, or a board you already have — no proprietary kit required.