Fluent.Demo.mp4
Watch the video above for a full walkthrough of how a practice interview turns into scored, actionable feedback.
An AI-powered interview practice platform for students and early-career professionals. Pick the role you're interviewing for, talk through a realistic interview out loud, and get graded feedback on your answers, structure, and delivery.
Practice is the core loop. You choose an interview scenario, or generate one from a real company and job title, and speak your answers out loud. An AI interviewer responds in real time with a natural voice, asks follow-ups, and pushes back the way a real recruiter or hiring manager would.
Analysis happens the moment the session ends. Every answer is scored across the dimensions that actually decide interviews: relevance, structure, specificity, and delivery. The result is an overall score plus a written breakdown of what worked and what didn't.
Coaching closes the loop. Fluent surfaces your weakest answer and rewrites it as a Perfect Response, so you see exactly how it should have sounded. Sessions are saved so you can watch your scores move over time.
Each scenario is a persona with a role, a company context, and a temperament. When you research a company, title, or interviewer, Fluent pulls public information into that persona so the questions match the job you're actually preparing for. The interviewer stays in character, never breaks to acknowledge it's an AI, and adapts its follow-ups based on what you just said.
| Step | Stage | What happens |
|---|---|---|
| 1 | Prepare | Enter the company, role, and interviewer, or pick from the scenario library |
| 2 | Practice | Speak through a live voice interview with an AI persona |
| 3 | Analyze | The session is scored across relevance, structure, specificity, and delivery |
| 4 | Review | Read the summary, strengths, and focus areas next to the full transcript and recording |
| 5 | Retry | Run the same scenario again and compare scores |
Step 5 is the point. One session tells you where you stand; five sessions is what actually changes the outcome.
- Live voice interviews — speak your answers and hear the interviewer respond with natural text-to-speech
- Scenario library — recruiter screens, behavioral rounds, hiring manager conversations, panel intros, and salary negotiations
- AI scenario generator — describe the job you're targeting and Fluent builds a custom interview at Easy or Hard difficulty
- Company and interviewer research — public context pulled in so the questions feel specific to your actual interview
- Post-interview analytics — overall score, per-dimension metrics, summary, strengths, and focus areas
- Transcript and playback — replay the real recording of the session alongside the transcript
- Perfect Response — an AI rewrite of your weakest answer
- Progress tracking — every session saved, with coaching summaries that show how you're trending
React 18 + TypeScript + Vite, styled with Tailwind CSS and shadcn/ui, with Framer Motion for transitions and TanStack Query for data.
Postgres with Row Level Security for auth, sessions, and stored recordings.
Edge functions (Deno) handle everything the browser shouldn't: the interviewer chat model, text-to-speech, company and interviewer research, grading, and transactional email.
Web Audio + MediaRecorder mix your microphone and the interviewer's voice into a single recording so playback sounds like a real conversation.
Fluent is audio-first by design. Interviews happen by voice, and feedback is grounded in what you said and how you said it. Video is not part of the current product.
The repository contains the full application: frontend, database schema and policies, and every edge function. Server-side keys for the AI, speech, research, and email providers are configured as backend secrets and are not committed.