InterviewLab

• 

• 

Inspiration

As an international student chasing the American dream, the job hunt felt like an uphill battle. I didn’t grow up with the same networks, cultural cues, or insider tips that many of my peers had. Every interview felt like a test not just of my skills, but of my ability to decode expectations I was never taught. I prepared the only way I knew how — with endless flashcards, generic advice, and late-night Google rabbit holes. This was before the AI agent boom we see today. Back then, there were no smart copilots, no personalized interview coaches — just forums, scattered blogs, and luck. And for someone trying to break into a new country and a new industry, it just wasn’t enough.

That’s why I built InterviewLab.

I didn’t have a tool like this when I needed it most — a coach that actually gets you. One that knows your domain, your role, and your ambition. One that doesn’t just quiz you, but trains you — like a mentor who’s walked the path before you.

Thanks to Bolt.new, I could move from idea to MVP at a speed I never imagined. Its zero-config dev environment, instant deploys, and developer-first design meant I could stop wrestling with setup and start solving the real problem.

To power the brains of InterviewLab, I turned to Perplexity’s Sonar API. Its deep contextual reasoning and ability to deliver nuanced, relevant feedback unlocked our vision: a smart, interactive coach built for the real world.

Bolt made it fast to build. Sonar made it worth building. But it was my own struggle — and the thousands like me — that made it necessary.

Now, InterviewLab is helping others like me prepare with confidence, insight, and personalized coaching. Because the dream is hard enough. Prep shouldn’t be.

What it does

InterviewLab harnesses the power of Perplexity's Sonar API to deliver an AI-powered interview simulation platform with features that weren't possible before:

Industry-Specific Intelligence: Sonar's knowledge base allows InterviewLab to understand the nuances of different roles and industries, providing relevant feedback for software engineers, product managers, data scientists, and more. Contextual Evaluation: The API evaluates answers based on their relevance to the specific question and role context, not just generic communication quality. Sophisticated Feedback Generation: Sonar analyzes responses across multiple dimensions (content, structure, relevance, communication) and generates specific, actionable improvement suggestions. Customizable Experience: Users select their target role, interview type, and preferred input mode (text or video), with the Sonar API adapting its evaluation criteria accordingly. Professional Voice Narration: Questions are presented with professional American female voice narration while Sonar's advanced language models work behind the scenes to evaluate responses.

How we built it

InterviewLab was architected and launched on Bolt.new — enabling rapid prototyping, instant deploys, and frictionless iteration from day one. Bolt’s lightning-fast environment let us stay focused on building the core AI experience instead of wrestling with setup or infrastructure.

At the heart of InterviewLab is the Perplexity Sonar API, specifically Llama 3.1-sonar-small-128k-online, which powers our intelligent evaluation engine.

Core Intelligence: Sonar analyzes user responses and delivers detailed, context-aware feedback across technical, behavioral, and case interviews.

API Integration: A custom middleware layer fine-tunes prompts and handles API communication, ensuring consistent output across scenarios.

Prompt Engineering: We crafted targeted, role-specific prompts that guide Sonar to evaluate based on real-world hiring criteria and produce actionable insights.

Frontend Framework: Built with React + TypeScript on Bolt.new, styled using Tailwind CSS and ShadCN UI for an accessible, responsive UX.

Challenges we ran into

Building with cutting-edge AI on a fast-moving platform like Bolt.new introduced exciting (and tough) engineering problems:

Optimizing Sonar Prompts: It took deep experimentation to extract consistent, high-quality feedback from the API across varied roles and question types.

Balancing Detail and Clarity: We had to distill Sonar’s rich feedback into something useful but not overwhelming — high signal, low noise.

Handling Diverse Answer Styles: Technical, behavioral, product-focused — every answer style needed different evaluation logic. Our prompts had to flex accordingly.

Tone Calibration: Encouraging but honest. We refined prompt tone to ensure users felt coached, not critiqued.

UI Integration: Parsing Sonar's unstructured feedback into clean, structured components in the frontend required precise formatting and rendering logic.

Accomplishments that we're proud of

Built rapidly on Bolt.new, our integration with Perplexity’s Sonar API unlocked several key advancements:

• Nuanced Feedback System: Leveraging Sonar’s deep contextual understanding, we crafted a feedback engine that delivers specific, actionable guidance tailored to each answer and context.

• Role-Specific Evaluation: By tapping into Sonar’s domain knowledge, we evaluate responses based on role requirements — differentiating what makes a strong answer for a software engineer versus a product manager.

• Multi-Dimensional Scoring: We designed a scoring model that assesses answers across multiple dimensions — content, structure, communication, and relevance — providing targeted feedback for each.

• Seamless Integration: While handling complex API interactions behind the scenes, the frontend built on Bolt.new offers users a smooth, intuitive experience.

• Optimized Prompts: Through extensive testing and iteration, we developed prompts that consistently extract high-quality, relevant feedback from Sonar across diverse interview scenarios.

What we learned

Building InterviewLab rapidly on Bolt.new and integrating Perplexity’s Sonar API taught us invaluable lessons about creating AI-powered products:

• API Capabilities: Sonar’s impressive depth across professional domains enabled us to deliver contextually rich, nuanced feedback — elevating interview prep beyond generic advice.

• Prompt Engineering Mastery: We refined the art of crafting precise, role-specific prompts that consistently drive high-quality, relevant AI outputs.

• Context Management: Learning to provide just the right amount of context to Sonar was critical to maintain response quality without overwhelming the model.

• Response Parsing: We developed efficient ways to extract and structure valuable insights from Sonar’s complex, detailed responses for a clear user experience.

• Designing AI UX on Bolt.new: Leveraging Bolt.new’s fast iteration and deployment, we experimented extensively with how to present AI feedback so it feels helpful, natural, and user-centered — a key factor in user engagement and trust.

What's next for InterviewLab

Enhanced Analysis: Leveraging more of Sonar's capabilities to provide even deeper insights into answer quality, including sentiment analysis and confidence assessment.

Interactive Coaching: Creating a more conversational experience where users can ask follow-up questions about their feedback directly to the Sonar-powered coach.

Expanded Industry Coverage: Adding specialized modules for more industries and roles, taking advantage of Sonar's broad knowledge base.

Comparative Analysis: Implementing features that compare user responses to ideal answers generated by Sonar, highlighting specific improvement opportunities.

Progress Tracking: Developing systems that use Sonar to track improvement in specific skills over time, providing personalized learning paths.

Mock Interview Scenarios: Creating more complex interview simulations with dynamic question selection based on user performance, powered by Sonar's reasoning capabilities.

InterviewLab demonstrates the transformative potential of Perplexity's Sonar API in creating personalized, intelligent learning experiences that adapt to each user's unique needs and context. We believe this application showcases just the beginning of what's possible with this powerful technology.

Built With

• css
• node.js
• npm
• shadcn
• sonar
• tailwind
• tanshack
• typescript