AI can write the code.
We assess the engineers who can build with it_
AI-Native Tasks are built from the start around AI-assisted engineering in a real VS Code environment, where using the AI Assistant is core to the task and treated as an engineering skill that separates strong engineers now.
Two kinds of task
Hiring teams now need to test two distinct things: can a candidate build AI products, and can a candidate work effectively alongside an AI Assistant, including knowing when to doubt it. The 10 VS Code tasks live today cover both: three build AI, seven work with AI.
Build AI
The candidate builds the AI system itself, and we score how they design and harden it. A grounded retrieval assistant, or a statistically sound anomaly detector.
Work with AI
The candidate works alongside an assistant to debug, investigate, or create. We score how well they direct it and whether they catch it when it is wrong.
Build AI
3 tasks liveThe AI or ML system is the deliverable. The candidate is judged on standing up a working product.
Legal Document Q&A Assistant
Implement a retrieval pipeline that answers natural-language questions about a legal document archive. It returns a generated answer alongside the source chunks it retrieved, so grounding can be checked.
DataSentinel
Build an anomaly-detection pipeline that ingests daily KPI readings, persists history to a database, detects statistical anomalies, and emits a structured report. Solvable in the candidate’s language of choice.
AI Model Churn
Build a churn prediction model for a language learning platform, covering feature engineering, model selection, and evaluation of how well the model actually predicts.
Work with AI
7 tasks liveA realistic engineering job alongside an assistant. The signal is how well the candidate directs it, and whether they catch it when it is wrong.
Order Router Microservice
Inherit a failing Python microservice, diagnose three latent production bugs, and get the test suite green. Maintenance work on code the candidate did not write.
Meeting Planner
Fix subtle bugs in an existing meeting planner, then build a new feature that suggests group time slots.
Log Analyzer
Build a tool that pulls incidents and patterns out of noisy, semi-structured logs and produces structured diagnostics.
Chat Conversation Analyzer
Build a command-line tool that explores chat logs and extracts intent, topics, and participants from messy dialogue.
Gallery Wall Curator
Solve an optimization problem, choosing items to maximize total value, designing and iterating heuristics with the assistant toward value thresholds.
React Carousel Music Player
Build a React carousel music player with audio playback, track metadata, cover art, play and pause, next track, and progress scrubbing.
Task Scheduler
Implement a scheduler API with dependency management, cycle detection, topological ordering, and critical-path scheduling.
Companion tasks, AI optional
The library also carries project-style VS Code tasks where the AI Assistant is available but not required. Realistic, multi-file engineering work that stands on its own with or without an assistant enabled.
Scored against the Engineering Skills Model
Every task is tagged against the Codility Engineering Skills Model at the test-case level, so each task reports against the specific skills its checks exercise, for example prompt engineering and AI integration for the retrieval assistant, or AI output validation and critical thinking for the debug and investigate tasks. Recruiters get the same skills-based reporting as the rest of the library, mapped onto role profiles. Validated by engineering leaders.
The set runs 40 to 80 minutes across easy, medium, and hard, covering back-end, front-end, data engineering, and data science profiles, so the same library serves early screening and senior interviews.