Skip to main content

Challenges For companies

// Challenge library

Challenges

Complete a challenge to add a verified artifact to your public profile. Build in any language — evaluated on behaviour, not syntax.

RAG Pipelines AI Agents MCP Servers Coding Agents Evals & Testing AI Tool Proficiency

Difficultyeasy medium hard

// 12 challenges · medium

Build a browser automation agent

Build an agent that completes browser-like workflows by navigating pages, reading state, filling forms, and submitting the correct result.

~4–8 h·Any language

Build a coding agent for data pipeline reliability

Build a coding agent that hardens a flaky data pipeline by adding deduplication, validation, idempotency, and clear processing counters.

© 2026 TryCrucible

~4–8 h·Any language

View →

Build a coding agent that writes tests for untested code

Build a coding agent that generates test suites for untested Python functions, covering happy paths, edge cases, and error conditions. The target functions are Python — your agent can be written in any language.

~4–8 h·Any language

Build a multi-hop RAG system for policy questions

Build a RAG system that answers policy questions requiring evidence from multiple documents.

~4–8 h·Any language

Build a RAG pipeline over multi-format documents

Build a RAG pipeline that handles mixed document formats — PDF text, markdown, and HTML — from a law firm knowledge base.

~4–8 h·Any language

Build a read-only SQL MCP server

Build an MCP server that exposes safe read-only SQL database tools for schema inspection and SELECT queries.

~4–8 h·Any language

AI Tool Proficiency

Build an admin dashboard feature using AI tools

Use Claude Code, Cursor, or GitHub Copilot to build a complete admin dashboard feature from a product specification.

~4–8 h·Any language

Build an AI agent for calendar scheduling

Build an AI scheduling agent that finds valid meeting times from attendee calendars and constraints.

~4–8 h·Any language

Build an AI agent that triages GitHub issues

Build an AI agent that automatically triages incoming GitHub issues: labels them, assigns priority, and drafts a first response.

~4–8 h·Any language

Build an MCP server for filesystem access

Build a Model Context Protocol (MCP) server that exposes a filesystem interface — allowing an AI model to list, read, write, and search files through standardised MCP tool calls.

~4–8 h·Any language

AI Tool Proficiency

Refactor a legacy Python codebase using AI tools

Use Claude Code, Cursor, or GitHub Copilot to refactor a messy legacy Python codebase. The target codebase is Python — your orchestration and tooling can use any language.

~4–8 h·Any language

Evals & Testing

Write evals for a text sentiment classifier

Design an evaluation suite for a sentiment classifier (positive/negative/neutral) that surfaces known failure modes: sarcasm, mixed sentiment, and domain-specific language.

~4–8 h·Any language