Intermediate everything in Backend Dev
NVIDIA's AVO harness ran Claude Opus 5 autonomously for seven days straight
NVIDIA paired Claude Opus 5 with a new harness called AVO to run long-horizon autonomous tasks, including a seven-day GPU kernel optimisation run and a separate reasoning benchmark. AVO uses persistent memory and a supervisor process so the agent keeps working past a single context window instead of restarting.
AI code volume is pushing code review earlier, ahead of the pull request
AI now produces more code than humans can realistically review, with Meta's lines of code per human-landed diff reportedly up 106% in a year. One response argues review should move earlier, into pairing and design sessions, leaving the pull request to catch formatting and known security issues.
PM2 cluster mode and multi-core utilization?
Cluster mode forks multiple worker processes using Node.js cluster module, distributes load, uses all CPU cores.
dependencies vs devDependencies in production?
Dependencies are needed at runtime, devDependencies only for development. npm install includes both, npm ci --only=production excludes dev.
What are key Dockerfile steps for Node.js Express apps?
Use lightweight base image, copy app, install deps, expose port, set NODE_ENV, run.
How to sync state across multiple user devices?
Attach userId to each socket, track active sockets per user, broadcast user state to all their sockets.
How to authenticate WebSocket connections using JWTs?
Client sends JWT on connection, server validates via middleware, socket is attached to user.
How does Socket.IO handle WebSocket failures with fallbacks?
Transports are communication protocols, WebSocket is primary, long-polling is fallback, client auto-detects and downgrades.
How do Socket.IO rooms organize client communication?
Rooms group sockets for targeted broadcasting, socket can join multiple rooms, emit to room broadcasts to all members.
Choosing Uvicorn worker count in production
Workers exist to use multiple CPU cores past the GIL; a common starting point ties count to cores, then you tune by load testing, balancing CPU and memory (each worker is a full copy) against connection-pool…
Reading the full response body in middleware
Responses stream as multiple body messages and headers go first, so you cannot add a header after seeing the body; you must buffer all chunks, compute the hash, set the header, then resend.
OAuth2 social login with your own JWT
A login endpoint redirects to the provider with a state param, a callback exchanges the code for the provider token, you fetch the user profile, upsert the local user, then mint your own JWT.
selectinload vs joinedload for eager loading
Use eager loading options to avoid lazy N+1; joinedload uses a single JOIN (good for many-to-one) but can fan out rows on collections; selectinload issues a second IN query (better for one-to-many).
Feature-based vs layer-based project structure
Layer-based groups by technical role and is simple early but scatters a feature across folders; feature-based groups by domain, improving cohesion and ownership at the cost of some duplication and…
Streaming large file downloads efficiently
Use StreamingResponse with a generator that yields chunks (or FileResponse for an on-disk file), set media_type and a Content-Disposition header, so memory stays flat regardless of file size.
Pydantic computed fields in response models
A @computed_field decorated property is excluded from input and validation but included in serialization and the OpenAPI schema, ideal for values like full_name derived from other fields.
Mapping camelCase JSON to snake_case Pydantic fields
Set an alias_generator (to_camel) plus populate_by_name in model config, accept aliases on input, and serialize with by_alias=True so responses come out camelCase.
Pydantic BaseModel vs dataclasses in FastAPI
BaseModel validates and coerces data at runtime, parses and serializes JSON, integrates with OpenAPI schema generation, and supports rich validators; dataclasses only store data with no validation.
Testing a DB endpoint via dependency override
Use app.dependency_overrides to swap the real get_db for one yielding a test database session, run against a disposable SQLite or test Postgres, and assert through TestClient.
Secure refresh token flow for access renewal
Short-lived access token, longer-lived refresh token stored server-side, a refresh endpoint that validates and rotates the refresh token issuing a new pair.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles