The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
AgentPrep AI — Real-Time AI Interview Coaching Platform
Turning Interview Practice Into a Structured, Personalized Coaching Experience
AgentPrep AI is an AI-powered interview preparation platform designed to help candidates practice realistic interviews, receive personalized feedback, and track their improvement over time.
Instead of relying on static question banks, articles, or repetitive self-practice, the platform creates an interactive interview environment where users can listen, respond, receive feedback, and improve.
The Challenge
Traditional interview preparation has several limitations:
Limited realism: Reading questions doesn't replicate an actual interview.
Inconsistent practice: Candidates often prepare without a structured routine.
Lack of feedback: Practicing alone makes it difficult to identify communication weaknesses.
Generic preparation: Standard question lists don't always reflect an individual's goals or target role.
No measurable progression: Candidates often have no clear way to understand whether their performance is improving.
AgentPrep AI was designed to bring these elements into one guided experience.
The Solution
We built an AI-powered interview coaching platform centered around three core principles:
Realistic Practice → Immediate Feedback → Continuous Improvement
The experience takes candidates from interview setup through live voice interaction and post-session analysis, creating a repeatable preparation workflow.
AI-Powered Mock Interviews
AgentPrep AI simulates interview scenarios using dynamic questioning rather than relying solely on static question lists.
Users can participate in structured mock interviews that encourage them to think and respond naturally under realistic conditions.
The experience is designed to replicate the conversational nature of an actual interview while maintaining a controlled environment for practice.
Voice-Based Interview Experience
Because real interviews are primarily verbal, the platform allows users to answer questions using their voice.
Whisper AI converts spoken responses into text, enabling the system to analyze each answer while allowing candidates to practice:
Verbal delivery
Response structure
Pacing
Clarity
Communication confidence
This creates a more realistic preparation experience than text-based practice alone.
Personalized AI Feedback
After each response, the platform analyzes the candidate's answer and provides personalized coaching feedback.
Claude Sonnet 4.5 powers the analysis layer, helping identify:
What worked well
Areas that need improvement
Communication patterns
Answer quality
Opportunities to strengthen future responses
Each answer becomes an opportunity to learn rather than simply another completed question.
Adaptive Preparation
The platform supports personalized question sets and learning paths based on the candidate's preparation needs.
Instead of repeatedly working through the same generic questions, users can follow a more focused preparation experience that evolves as they continue practicing.
This creates a feedback loop where:
Practice → Analysis → Improvement → More Practice
Performance Dashboard
We also developed a performance analytics experience that allows users to monitor their development over time.
The dashboard helps candidates understand:
Practice activity
Performance trends
Areas requiring additional attention
Progress across sessions
Overall preparation consistency
This adds a measurable coaching layer to the interview simulation experience.
Product Experience
Interview Introduction
A guided starting screen explains the AI interview process and prepares users for the upcoming session.
Interview Length Selection
Users can select a shorter or longer practice session depending on their available time and preparation goals.
Live Interview
Candidates receive questions and respond naturally through voice, creating a focused environment that closely resembles a live interview.
Feedback
Once a response is complete, AI-generated coaching highlights strengths and provides specific areas for improvement.
What We Delivered
The completed platform included:
AI-powered mock interviews
Dynamic interview questioning
Voice-based responses
AI transcription
Personalized answer analysis
Adaptive preparation flows
Performance tracking
Interview session management
Progress analytics
Responsive web experience
Results
AgentPrep AI transformed interview preparation from an unstructured activity into a guided coaching workflow.
The platform helped users:
Prepare 75% more efficiently
Practice speaking rather than simply reading answers
Receive immediate, personalized feedback
Identify recurring weaknesses
Track improvement across sessions
Build a more consistent preparation routine
The result was an interactive preparation system designed to help candidates move from practice to measurable readiness.
Technology Stack
Frontend: Next.js, Tailwind CSS
AI & Voice: Whisper AI, Anthropic Claude Sonnet 4.5
Core Experience: AI interview simulation, voice interaction, personalized feedback, adaptive coaching, performance analytics
The Result
AgentPrep AI turned interview preparation into an active, feedback-driven learning experience.
Rather than simply providing candidates with more interview questions, the platform gives them a place to practice realistically, understand their performance, and continuously improve.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
no-code.supply is open.
I've been quietly building landing page templates. Six are out today, each with its own brand, its own type pairing and one interaction it's built around:
→ Aurel: a skincare brand with a WebGL serum drop that bulges toward your cursor
→ Ferro Type: a type foundry where the hero word bends its weight and width under the pointer
→ Meridian Expeditions: contour maps that redraw themselves as you switch routes
→ Oda: an architecture studio whose timber cabin draws itself, with a live sun study
→ Relay: a dev tool where the editor types a workflow and the run graph executes it live
→ Signal: a mission-control SaaS with a live infrastructure graph that contains threats as they flare
Built to be edited with AI. Every template ships with a README, a DESIGN.md and an AGENTS.md, so Claude, Cursor or whatever assistant you use changes it the way it was built, not the way it guesses.
Pricing, per template:
• HTML: $29
• React: $59
• Bundle (HTML + React): $79
Or get All-Access for $169: all six templates, both editions, plus every future drop. Nine more are on the way, at no extra cost.
No page builder, no lock-in, no third-party requests.
If you build sites for clients, I'd love to hear which one you'd use first.
Ferro Type is the one I would open first. Bending a variable font under the pointer re-lays out the text every frame, which is a much bigger bill than it looks, and stepping the weight in a few fixed increments is usually what keeps it smooth on a laptop that is not plugged in....
GROOOMY 🐶 Smart booking for a one-woman dog groomer
Grooomy is a booking app for Marloes, who runs a dog grooming atelier on her own. Customers book in about a minute, and the app reserves exactly the table time each dog needs.
What I built: Every groom takes a different amount of time. A Chihuahua takes an hour, a matted Bernese takes four. Marloes used to plan every appointment over WhatsApp: nine messages, two days and the occasional no-show. Grooomy fixes that:✅
✅ Customers describe their dog (size, coat, tangles), and the duration and price are calculated automatically.
✅ Only time slots where the full groom fits are shown, so there are no double bookings and no gaps too short to use.
✅ Returning customers are recognised by phone number, and new customers pay a small deposit.
✅ Customers can change or cancel with one tap.
✅ When a day is full, dogs join a waitlist, and a cancellation is offered to the next matching dog automatically.
✅ Marloes gets a dashboard with her day at a glance, the week, the waitlist, customers and every automated message.
How I built it: I started with the scenario and design research, designed the brand and every screen, then turned each screen into a precise prompt for Lovable. The app runs on Lovable Cloud with a real database for bookings, customers and the waitlist. The design process video is on the Story page.
CHECK IT OUT: https://grooomy.lovable.app
___________
Worked really hard on this and loved working with Lovable for the first time! Let me know what you think! 🔥
I've never made an ad like this and I'm really proud of it! It's for Cosign, an app I designed that helps users understand and manage terms and agentic permissions.
Digital consent, basically.
Every frame rendered in Blender on my M1 Max, which took a while.
Figma x Opus 5.5 x Blender
