# Custom technical challenge authoring

Intervieux lets employers write their own technical challenges, coding, debugging, SQL, or written, with AI help drafting both the problem and a suggested answer key, and a test-run available before the challenge ever reaches a real candidate.

A generic coding challenge tells you whether someone can code. A challenge built around the actual problems your engineers hit tells you whether someone can do the job. Custom authoring exists so a technical challenge can reflect the real work a role involves, without requiring whoever writes it to also build a full answer key and test cases by hand from nothing.

## What it does

Challenge authoring covers the full lifecycle of a technical assessment: writing the problem itself, defining what a correct or strong solution looks like, and verifying the challenge actually works before a candidate sits down with it. Challenges can be coding, debugging, SQL, or written format, and coding challenges run on real code execution with language auto-detection, so a candidate's actual submitted code gets run and checked, not just read.

## How it works

1. **Start with an AI-drafted problem** — Describe the kind of challenge you want, a coding problem, a debugging scenario, a SQL query task, and AI drafts a first version of the problem statement to start from.
2. **Get a suggested answer key** — Alongside the problem, AI suggests an answer key, giving a baseline for what a correct solution looks like before a human has to write one from scratch.
3. **Set test cases and difficulty** — Hidden and visible test cases can be attached to check a candidate's actual submission, along with hints and a difficulty level appropriate for the role's seniority.
4. **Test-run it yourself before publishing** — Before a challenge goes live for candidates, it can be test-run directly, catching a broken test case or an ambiguous problem statement before it ever reaches someone being evaluated on it.
5. **Candidates submit real code that actually runs** — When a candidate takes the challenge, coding submissions execute for real through the code execution engine, with language automatically detected across common languages, rather than the answer being scored on appearance alone.

## Real execution, not pattern matching

Coding submissions run on an actual execution engine covering JavaScript, Python, and Java directly, with auto-detection extending across additional languages including TypeScript, C++, Go, Rust, and C#. A candidate's code is checked against real test cases, not scanned for keywords.

## Hidden test cases matter for integrity

Visible test cases let a candidate check their own work as they go. Hidden test cases run behind the scenes and catch a solution that only handles the example shown, which is the difference between a candidate who solved the actual problem and one who pattern-matched the visible case.

## Difficulty and hints let one challenge fit different levels

The same underlying problem type can be configured with a difficulty level and optional hints appropriate to who is taking it, a junior candidate and a senior candidate can face meaningfully different versions of a related challenge rather than the exact same fixed test.

## Challenge results feed the rest of the pipeline

A technical challenge produces a score that lives on the candidate's card alongside their five AI interview dimensions, and that challenge score is one of the fields the screening rules engine can evaluate directly, so a custom challenge built for a specific role can also drive automated tagging or stage moves once candidates start submitting.

## Who this is for

This is for hiring teams filling technical roles where a generic off-the-shelf coding test does not reflect the work, an engineering team wanting a debugging scenario pulled from a real class of bug they see, or a data role wanting a SQL challenge that mirrors the actual queries the job involves.

## Frequently asked questions

### What formats can a custom technical challenge take?

Coding, debugging, SQL, and written formats are all supported, so the challenge type can match what the role actually requires rather than defaulting to a single coding-only format.

### Do I have to write the problem and answer key myself?

No. AI can draft the problem statement and suggest an answer key as a starting point, which can then be edited before the challenge is published.

### Can I check that a challenge works before candidates see it?

Yes. A test-run is available before publishing, so a broken test case or unclear problem statement can be caught and fixed ahead of time rather than discovered mid-interview by a real candidate.

### How is a candidate's coding submission actually evaluated?

Coding submissions run on a real code execution engine covering JavaScript, Python, and Java, with language auto-detection extending to additional common languages, checked against hidden and visible test cases rather than scored on appearance.

### Does a technical challenge result affect anything beyond the challenge itself?

Yes. The challenge score sits on the candidate's card alongside their AI interview scores, and it is one of the fields the screening rules engine can evaluate, so it can drive automated tagging or stage changes just like the interview scores can.

## Related pages

- [Technical AI interviews](/features/technical-interviews)
- [AI-drafted job setup and scoring rubrics](/features/ai-job-setup)
- [Reusable interview templates](/features/interview-templates)
- [Compare Intervieux to Karat](/compare/intervieux-vs-karat)
- [Frequently asked questions](/faq)

## Write a technical challenge that reflects the real work

Draft a problem with AI help, get a suggested answer key, and test-run it yourself before it reaches a single candidate.

Start practicing free: https://www.intervieux.ai/register · Hire with Intervieux: https://www.intervieux.ai/employers/signup