How Vivasoft Used AI to Fix Interview Prep for Job Seekers
Project Type
AI Product
User
Job Seekers
Industry
Recruitment Technology
Team Size
7
Service Used
Engagement Length
December 2025 - Ongoing
About Previea
Previea is an AI-powered mock interview platform developed by Vivasoft. It runs live, job-specific interviews for candidates and scores their answers automatically.
Previea is a tool that lets a candidate practice a real interview and get feedback on their answers, without needing another person to conduct it.
Project Overview
Job seekers often have limited ways to practice for a real interview. A mock interview usually needs a second person, and that person is not always available. Even when practice happens, feedback is rarely objective or immediate, and most practice questions are generic rather than matched to the candidate’s actual target role.
Vivasoft decided to build a practice tool that could conduct and evaluate mock interviews. The team designed an AI interviewer that runs a live interview session from start to finish. Candidates choose their job or subject, upload their CV, and complete a device and consent check first. The AI asks relevant questions, listens through speech-to-text, and evaluates each response as the interview happens.
After the session, Previea generates a report with scores and feedback. To keep sessions fair, Vivasoft added monitoring for tab switching, fullscreen exit, and camera activity. Previea launched in Bangladesh and now runs 250 to 300 interviews a day.
Technologies Used
Vivasoft’s AI engineering team used the tools listed below to meet Previea’s real-time, fairness, and scaling requirements, supporting each part of the platform from speech processing to security and payments.
Python
Go
JavaScript
FastAPI
Echo
React.js
PostgreSQL
AWS
Claude
Google Gemini
Whisper
JWT
SSLCommerz
Challenges Faced by Vivasoft
Building Previea required the team to develop a live AI interviewer that could conduct an interview from start to finish. The team had to solve for realism, fairness, and scale at the same time. Here are the challenges the development team faced while building Previea:
1. Building a Natural, Real-Time AI Interview
The AI needed to conduct a live interview while listening, processing responses, and generating the next question without noticeable delays. Vivasoft tuned tone, pacing, and question sequencing across many test rounds.
2. Generating Job-Specific Questions
Interview questions had to fit each candidate’s exact job or subject, not come from a fixed bank of generic questions. The AI needed to generate relevant questions on the fly, based on that specific input.
3. Scoring Answers Consistently
Candidate answers vary widely in length, phrasing, and detail across every session. The AI had to score each candidate against the same standard, regardless of how they answered or what they emphasized.
4. Managing Device Access and Cheating Detection
The platform had to manage camera and microphone permissions reliably across devices, recovering gracefully when a check failed. It also had to catch signals like tab switching or multiple faces without flagging genuine candidates.
5. Scaling for Concurrent, Accessible Use
The system had to stay responsive as many candidates ran interviews at once. It also had to stay simple for users with different levels of technical comfort, from CV upload to the final report.
Solutions by Vivasoft
Vivasoft addressed each challenge with a specific part of the interview flow, from the live AI session to the checks that run around it. Here are the solutions Vivasoft implemented during development:
1. Building a Real-Time, Adaptive Interview Flow
The AI team built the interviewer to process each candidate’s answer as it comes in, then generate the next question in real time rather than pull from a script. The team ran repeated test rounds, adjusting tone, pacing, and sequencing until the flow felt like a real conversation.
2. Connecting Questions to Each Candidate's Role
Vivasoft connected the question generator to each candidate’s selected job or subject and CV, so it builds a fresh set for every session instead of pulling from a fixed bank. The team tuned this step until questions stayed tied to the actual role, not just the general field.
3. Building Consistent Scoring Logic
The development team built the evaluation model to score every response against the same fixed criteria, regardless of how a candidate phrased their answer. Vivasoft tested this logic across many different answer styles and lengths, so scoring stayed consistent from one candidate to the next.
4. Layering Device Checks and Cheating Detection
The engineering team built the onboarding flow to request camera and microphone access upfront, with checks that catch device issues before the interview starts. Vivasoft layered in monitoring for tab switching, fullscreen exit, copy or paste, camera status, multiple faces, and looking away, tuned to flag suspicious behavior without interrupting genuine candidates.
5. Scaling the Backend and Simplifying the Flow
Vivasoft built the backend on FastAPI and Go so multiple interview sessions could run at once without slowing down. The team also simplified the interview flow step by step, from CV upload to calibration, so candidates with different levels of technical comfort could complete it without help.
Outcomes of the Project
Previea has been running since December 2025, and early usage gives Vivasoft a working, real-world read on the product. Here are the outcomes from the development and continued use of Previea:
Steady Daily Usage
Previea now sees 250 to 300 daily users across Bangladesh, a volume that has held steady and grown gradually since the platform’s launch in December 2025.
Feedback Right After the Session
Previea can now deliver the completed interview report within 30 to 60 seconds of the session ending, giving candidates feedback while the interview is still fresh in their minds.
Practice Without Scheduling a Person
Each user can complete one interview at a time, keeping the session focused from the first question through the final evaluation without overlapping interview sessions.
Fast AI Interview Turnaround
The AI interviewer responds within an average of 3 to 4 seconds after a candidate finishes an answer, keeping the interview moving with minimal waiting between questions.
Team Involvement
| Resources | Count |
|---|---|
| Front-End Developer | 1 |
| Back-End Developer | 2 |
| QA Tester | 1 |
| AI/ML Specialist | 2 |
| Product/Business Analyst | 1 |
Core Features of the Software
Previea is designed to conduct mock interviews and evaluate candidate responses. The features below cover the full flow, from onboarding to the final report.
Live, Personalized AI Interview
Candidates go through a real, conversational interview run entirely by AI, built around their chosen job or subject. The AI also uses the candidate’s CV to shape its questions.
Automated Evaluation and Instant Feedback
The AI evaluates each response as the interview happens and generates a structured performance report right after. Candidates see their strengths and improvement areas without any waiting period at all.
Guided Onboarding and Device Calibration
Candidates verify their identity through OTP before starting the interview. A pre-interview check then confirms their camera and microphone are working correctly, so the session runs smoothly from the start.
Anti-Cheating Monitoring
The platform detects tab switching, fullscreen exits, copy/paste, multiple faces, and looking away, all in one monitoring layer. The platform records these activities as part of the interview session.
Usage Tracking and Payment Integration
The platform tracks how many interviews a candidate has used or has remained at any given time. Additional sessions can be purchased through an integrated SSLCommerz payment gateway.
Multilingual, Responsive Experience
Candidates can take interviews in either English or Bangla, based on their own preference. The interview flow also works consistently across different screen sizes and devices used.
Development Timeline
Project Start Time
Dec 2025
Project End time
Ongoing
Ready to Build an AI-Driven Interview Platform?
