AI Enterprise SolutionsGPT & Claude

Quadra

Quadra is designed for job seekers preparing for technical and professional interviews, as well as recruitment teams that need a more structured way to assess candidates. Today, candidates often lack access to realistic interview practice, actionable feedback, and an affordable way to identify weaknesses before a real interview. Recruiters, meanwhile, spend significant time conducting repetitive initial interviews and organizing candidate responses. Quadra aims to bridge this gap through an AI-powered interview experience that asks role-specific questions, captures candidate responses, and helps turn interview practice into measurable improvement. The initial focus is technical job seekers and early-career candidates who need practical interview preparation without depending on expensive coaching or constant access to a human interviewer.

Open the demo Video GitHub

Preliminary score

47.0/ 100Earlyaverage of 2 judges

Value for the user14.5 / 25
A specific problem and a clear outcome from solving it.
Prototype and use of AI14.0 / 30
A working core scenario, and what AI actually contributes to the result.
Quality testing8.0 / 20
Tests, examples of failures, and comparison with the current approach.
Feasibility7.0 / 15
Data requirements, running costs and a clear next step.
Originality3.5 / 10
A meaningful difference from common solutions.

What the judges wrote

Claude Opus 5.5Preliminary46/100
Strengths
Your screenshots show a real two-sided product: a trilingual (AZ/EN/RU) candidate portal with live camera preview, timed role-specific questions, and a recruiter pipeline with per-candidate scores and status filters. Your README describes a substantial stack (React 19 + Vite, ASP.NET Core 9, role-based access, enforced subscription and AI usage limits, Docker Compose, CI), which suggests serious engineering effort. Your testing section honestly admits a concrete defect: an irrelevant role/skill prompt and leftover test text appeared in the conversation area, and face-detection and analytics states were mishandled when no data existed.
To improve
Your submission lists test scenarios but reports no results. Run them on a set of real role and answer pairs, and report relevance ratings, speech-recognition error rate and latency against a human-rated baseline. Several fields read like unfilled templates ('any additional models should be listed separately', 'the project's actual frontend and backend frameworks'). Name the exact models, speech and vision libraries, and the data each metric is computed from. The recruiter view shows candidates auto-rejected below a 70% threshold using 'stress' and 'eye contact' scores, which conflicts with your 'support, not replace' message. Validate these signals for fairness and keep a human in the loop. For the enterprise track, pick one hiring workflow, define an adoption metric such as recruiter hours saved per shortlist, and add per-interview AI costs and pricing to your annual cost estimate.
GPT-6 AstraPreliminary48/100
Strengths
Your combined candidate-practice and recruiter-review workflow addresses recognizable needs on both sides of hiring. Your deck includes concrete interface examples, and your README provides a specific technology stack and Docker setup instructions. Your testing notes usefully acknowledge irrelevant prompts, session-state problems and unreliable indicators when valid input is unavailable.
To improve
Your next demonstration should show a complete role-selection-to-feedback session and identify which model and speech integrations actually run. Report reproducible test cases with expected and actual outputs, including failures, and compare relevance and time spent against static question lists or human-led screening. Keep unvalidated eye-contact, stress and confidence indicators out of hiring scores, explicitly distinguish missing data from poor performance, and demonstrate human review. Supplement your annual budget with inference and hosting costs per interview, candidate-data consent and retention requirements, and a narrowly scoped pilot with a measurable adoption target.

The team

Turkan Karimova
product
Aytaj ShakarzadeLead
developer
Nurlana Sadiqli
data
Nargiz Mehdiyeva
developer

Disclosure

Required by the rules. Judges verify it against the repository.

Models: OpenAI GPT-4o mini for interview question generation and response analysis, if enabled in the current build. Any additional models used for speech recognition or visual analysis should be listed separately.
Data: Synthetic interview scenarios and user-provided responses. No private recruitment dataset is claimed unless one was actually used.
Components: The project's actual frontend and backend frameworks, AI API integrations, speech-processing libraries, and any database or hosting services used in the current implementation.
Built during the event: Confirmed

Quality testing

**Testing approach** We evaluated Quadra against common interview preparation methods using representative technical interview scenarios. **Test scenarios** 1. Relevant technical question — checked whether the AI interviewer generated a question aligned with the selected role. 2. Voice input — checked whether spoken answers were captured and processed correctly. 3. Language selection — checked whether the interview experience followed the selected language. 4. Missing camera input — checked whether the system handled an undetected face without presenting misleading visual indicators. 5. Invalid or irrelevant responses — checked how the system handled unexpected input and whether the conversation could continue. **Observed limitations** During prototype testing, the interface displayed an irrelevant role and skill prompt, while the conversation area contained test text. Analytics and face-detection states also require clearer handling when no valid data is available. These issues highlight the need for stronger input validation, more reliable session-state management, and explicit handling of unavailable measurements. **Comparison with the current approach** Traditional interview preparation often depends on self-practice, static question lists, or access to a human coach. Quadra aims to provide an interactive alternative with role-specific questions and a structured conversation history. **Next testing steps** Measure question relevance, speech recognition accuracy, response latency, session completion rate, and the consistency of feedback across a set of representative test cases. Results will be reported only after the tests are completed.

Worth 20 of the 100 points.

Running it

**How to test Quadra** 1. Open the live demo in a modern desktop browser. 2. Allow microphone access when prompted. 3. Select the interview language and begin a session. 4. Answer the AI interviewer's questions using your microphone or the available text input. 5. Review the conversation history and any analytics successfully generated during the session. **Testing notes** * Use a quiet environment and a working microphone for voice-based interviews. * Camera-dependent indicators require camera permission and a detectable face. * If a feature is unavailable or permission is denied, do not interpret missing data as a candidate performance score. **Known limitations** This is an early-stage prototype. Feature availability and analytical accuracy depend on the implemented integrations and the user's device permissions.

Project ownership

The team keeps the rights to this project. Judges were given access for verification only.