Overview

The AI Coding Interview feature provides an objective evaluation of a candidate's technical and communication skills. To do this, Eightfold uses a comprehensive scoring system that analyzes performance across several key areas.

Eightfold AI Interviewer supports the following coding languages:

• Javascript
• Typesript
• Python
• Java
• C++
• PHP
• Ruby
• Go
• Rust
• Swift

Scoring process

Eightfold uses a structured, five-step approach to turn raw interview data—like the final code and transcript—into actionable insights

1. Data gathering: The system collects all inputs for each question, including the submitted code, conversation transcript, and any hints provided by the AI agent.
2. Question-wise analysis: Each coding question is evaluated independently for a granular assessment.
3. Signal generation: The system scores the candidate's performance against specific criteria called signals. Each signal score includes reasoning and evidence from the interview.
4. Metric aggregation: Related signals are grouped into broader performance categories called metrics. These summarized insights are what you see in the feedback report.
5. Final score calculation: The metric scores are averaged to generate an overall score for each question.

Signals and metrics

The coding interview evaluation is built on signals (atomic insights) which are grouped into metrics (holistic skill categories). Here are the key metrics and the signals:

Code quality

This metric assesses the technical quality of the solution. It's a weighted average of three signals, which are measured against a predefined baseline for objectivity:

• Code correctness: Does the solution produce the correct output for all inputs, including critical variations?
• Time and space complexity: Did the candidate implement an efficient algorithm?
• Edge case handling: Did the candidate anticipate and correctly handle boundary inputs (for example, empty arrays, null values)?

Code readability

This metric evaluates the code's clarity and structure. It's a weighted average of:

• Code modularity: Is the code well-structured into reusable functions?
• Code hygiene and style: Is the code clear, with good naming and formatting?

Problem solving

This metric assesses the candidate's overall approach. It's a weighted average of:

• Hint reliance (Heavily weighted): How independently did the candidate solve the problem? Relying on more hints indicates less independent problem-solving.
• Debugging: How well did the candidate identify and resolve bugs?

Collaboration

This metric evaluates communication and responsiveness. It's a weighted average of:

• Communication: How well did the candidate articulate their approach and ask clarifying questions?
• Clarity of thinking: Was the candidate's explanation logical and easy to follow?
• Hint application: How effectively did the candidate apply hints? Candidates who require no hints score highly on this signal.

Feedback report

After the interview, you can access a detailed feedback report that includes:

• An overall interview score.
• Detailed explanations with positive/negative feedback
• Question-wise ratings for each metric.
• A "View Code" feature to review the candidate's final submitted code for each question.

This comprehensive feedback helps you scale your hiring process, reduce unconscious bias, and gain objective insights into a candidate's skills.