The Challenge
Manual Screening Bottleneck
Hiring at a small agency means the founder and key team members are directly involved in evaluating candidates. Without a structured system, screening involves reading resumes, conducting multiple rounds of manual assessment, and maintaining spreadsheets of candidate progress. For every open position, this can consume 20-30 hours of senior time.
No Standardized Assessment
Without a consistent test framework, every candidate gets evaluated differently depending on who interviews them and what questions come to mind. This makes comparing candidates across applications unreliable and introduces unconscious bias in the evaluation process.
Aptitude Test Integrity
Online assessments are vulnerable to cheating - tab switching to search for answers, copying questions to AI tools, sharing answers between candidates, or using browser automation to auto-complete tests. Any assessment platform that does not address these vectors produces unreliable scores.
Question Bank Creation
What We Did
Admin Panel
Job Management
The admin creates job postings with rich text descriptions (TipTap editor), configures department, location, job type (Full-Time or Intern), work location (Remote, On-Site, Hybrid), and publication status (Draft, Published, Closed). Each job has per-test-type configuration: question count, time limit in seconds, and minimum pass threshold percentage. Custom application fields can be added per job (e.g., GitHub URL for developers, Behance for designers) with type, label, validation rules, and required/optional flags.
MBTI Suitability Profile
For each job, the admin answers a 10-question scenario-based suitability questionnaire. Each scenario maps responses to MBTI dimension preferences (E/I, S/N, T/F, J/P). The system aggregates these responses to determine which MBTI personality types are most suitable for the role, creating a data-driven personality-job fit profile rather than relying on gut feel.
AI-Powered Question Generation
The admin panel includes bulk AI question generation via OpenRouter API (GPT-4o-mini model). Questions can be generated for all 6 test types with tailored prompts per category. Batch generation supports up to 100 questions per request with automatic retry for shortfalls. An intelligent duplicate detection system uses normalized text comparison - handling word order variations, location differences, filler words, and key phrase preservation - to prevent duplicates from entering the bank. The current production bank contains approximately 418 validated, production-ready questions.
Candidate Management
The Results
Complete Product Build Capability
This is not a form builder with a quiz plugin. It is a full-stack recruitment platform with 12 database models, 17 API endpoints, AI integration, anti-cheating infrastructure, file storage, transactional email, and a complete admin panel. Built as a production system, not a prototype.
AI-Powered Content Generation
The question bank generation system demonstrates practical AI integration - using GPT-4o-mini through OpenRouter to generate validated assessment questions at scale, with intelligent duplicate detection ensuring bank quality. This same pattern applies to any domain where AI-generated content needs to be produced, validated, and managed.
Security-Conscious Architecture
Key Takeaway
Complete Product Build Capability
This is not a form builder with a quiz plugin. It is a full-stack recruitment platform with 12 database models, 17 API endpoints, AI integration, anti-cheating infrastructure, file storage, transactional email, and a complete admin panel. Built as a production system, not a prototype.
AI-Powered Content Generation
The question bank generation system demonstrates practical AI integration - using GPT-4o-mini through OpenRouter to generate validated assessment questions at scale, with intelligent duplicate detection ensuring bank quality. This same pattern applies to any domain where AI-generated content needs to be produced, validated, and managed.
Security-Conscious Architecture
