Python backend interviews cover language internals, async programming, Django or FastAPI, and database access patterns. Below are the most common questions with model answers. Junior: core theory, definitions, and simple practical cases.
The Global Interpreter Lock allows only one thread to execute Python bytecode at a time in CPython, so threads never run Python code in parallel across cores. Threads still help for I/O-bound work because the GIL is released during blocking calls. For CPU-bound work you need multiprocessing, a C extension that releases the GIL, or a different runtime.
Lists are mutable and tuples are immutable, which makes tuples hashable and usable as dictionary keys. Tuples are slightly faster and signal to the reader that the collection is a fixed record rather than a growing sequence. Use a list when the contents change and a tuple for fixed heterogeneous data such as coordinates or a returned pair.
A decorator is a callable that takes a function and returns a replacement, used to add behaviour without touching the original body. A timing decorator wraps the call: record perf_counter before, invoke the function, record after, log the delta, and return the result. Always apply functools.wraps to the inner function so the original name, docstring, and signature survive introspection.
A regular function computes everything and returns once; a generator suspends at each yield and resumes where it left off, producing values lazily. That means constant memory regardless of how many items are produced, so you can iterate a multi-gigabyte file or an infinite sequence. yield is what turns the function into a state machine the interpreter can pause and resume.
*args collects extra positional arguments into a tuple and **kwargs collects extra keyword arguments into a dict. They are used when a function must accept an arbitrary signature — most often in decorators and wrappers that forward everything to the wrapped callable. The mirror use is unpacking: calling f(*items, **options) spreads a sequence and a mapping into arguments.
Threading gives concurrency within one process and helps only with I/O because of the GIL. Multiprocessing spawns separate interpreters with real parallelism, at the cost of process startup and inter-process communication — the right choice for CPU-bound work. asyncio is single-threaded cooperative concurrency: tasks yield at await points, which scales to tens of thousands of sockets but blocks everything if you call synchronous code inside a coroutine.
A context manager guarantees setup and teardown around a block, which is what with statements use to close files or release locks even when an exception is raised. You implement one either as a class with __enter__ and __exit__, or as a generator decorated with @contextlib.contextmanager that yields once between the setup and cleanup code. The __exit__ method receives exception details and can suppress the exception by returning True.
The Method Resolution Order defines the sequence in which Python searches classes for an attribute, computed by the C3 linearization algorithm. It guarantees that a class always precedes its parents and that the order of base classes is respected, which makes diamond inheritance predictable. super() follows the MRO rather than jumping straight to the parent, which is why cooperative multiple inheritance requires every class in the chain to call super().
N+1 occurs when a queryset returns N objects and each access to a related field triggers its own query. select_related solves it for ForeignKey and OneToOne by performing a SQL JOIN and populating the related object in the same query. prefetch_related handles ManyToMany and reverse relations by running a second query and joining the results in Python, which is why it produces two queries instead of one.
In a real interview you speak under pressure. Cam asks these same questions, scores every answer, and shows exactly what to fix.
Practice a Junior Python Backend interview →