InterviewForge

Logo
An Interview Call
Landing Page
An Interview Report
AI analysing your Job

Inspiration

Job interviews felt like a one-shot game—high stakes, no practice field. I wanted to give candidates a “time-machine” loop: rehearse, refine, repeat until ready. Advances in multimodal AI, browser APIs, and affordable CVI platforms finally made that dream realistic—so I built InterviewForge in one day.

What it does

Architecture

Turns any job post into a live mock interview. Paste a JD URL or text and hit Start.
Generates tailored questions. An LLM plus deep research pipeline mines role-specific, skill-level-appropriate questions and interview patterns.
Runs a full-screen video interview. Using Tavus and Daily for CVI, a super-realistic avatar comes on screen to interview you
Delivers instant feedback. We score each answer in each attempt for content, clarity, and confidence, cross-check facts, and flag knowledge gaps.
Analyzes body language. Tavus' cutting edge perception models pick up on visual cues and respond to them in the call, and presents them for analysis after.
Enables rapid iteration. One click relaunches another round or replays past attempts with annotated insights - focusing on parts that you can improve

How I built it

Frontend: React + Vite, Tailwind and shadcn-ui for styling, Tavus for CVI with the Daily React SDK
Backend: Supabase Serverless Functions paired with Supabase's DB offering - auto deployed from bolt.
AI Video Call engine: Tavus with multi-layer use (Perception, STT, TTS, LLM & Realtime Replica) . On the fly creation of personas for every single job description
DevOps and Deployment: Auto deploy to Netlify via Bolt
Auth: Supabase Auth using Supabase RLS to enable fine grained access control

Challenges I ran into

Extremely large bolt projects which needed to be rolled back and restored via backup
Hooking into daily events and the daily react SDK
Debugging serverless functions that fanned out to other serverless functions is a challenge without tools like opentelemetry

Accomplishments that I am proud of

Idea to production in <12H: I started building this on Jun 29th. Thanks to Bolt, Tavus and over 10M credits (Thanks StackBlitz team!), I was able to get this from an idea in my head to something you can sign up and use right now!
High Fidelity Questions: The deep research pipeline creates personas with a high quality knowledgebase of questions and role related understanding - blew me any my friends away!
Used in practice: 2 beta testers were able to actually use this to prepare for their interviews on the 30th - and they have glowing reviews!

What I learned

Comprehensive SDKs like the one from Tavus allow you to rapidly prototype. This experience would have been super hard to build without them.
It's an exhilarating experience to decide to take part in a 30 day hackathon on the very last day - but thanks to tools like Bolt - I've built something that I am genuinely stoked to launch as a Micro-SaaS ## What's next for InterviewForge

Pricing Model & Micro-SaaS launch

The app already has the fundamentals for the pricing model set - there's a Free/Pro tier. You can pay a specific amount for a certain number of attempts OR use it as a subscription service with a certain number of free attempts per month. Due to Stripe not operating in my country, I will have to set up alternatives to accept payments - which take approval time

Team Up with Job Platforms

Nobody's more interested in getting prepared candidates more than job platforms like LinkedIn. Partner with them to offer this as a white-label service for their premium tier subscriptions

Match By Interview

What if instead of just preparing you, InterviewForge could use your strengths gathered from all the interviews you've given to match you to jobs that you can likely get.