Inspiration
The origin of PocketInterview is rooted in the significant challenges faced by job seekers in Indonesia: a persistent high unemployment rate and a glaring mismatch between available skills and industry demands. Many talented individuals, despite their potential, struggle to secure employment due to a lack of effective interview practice and underdeveloped communication skills. Simultaneously, we recognized the burgeoning global job market, where remote work is increasingly common. This presents a massive opportunity for Indonesian talent, provided they can confidently navigate international interview processes and communicate effectively in English. PocketInterview was conceived to bridge this gap, empowering individuals to master their interview skills and unlock both local and global career opportunities.
What it does
PocketInterview is an AI-powered mobile application designed to provide realistic and personalized mock interview experiences. It allows users to practice with two distinct AI interviewers: "Steve" for technical interviews and "Lucy" for behavioral interviews.
Key features include:
- Personalized Interview Sessions: Users can upload their CVs (PDF, DOC, DOCX), which the app analyzes to tailor interview questions based on their specific skills, experience, and background.
- Comprehensive AI Feedback: After each session, users receive detailed scores on clarity, grammar, and the substance of their answers, accompanied by specific reasons for each score.
- Full Transcripts: Every conversation is transcribed, allowing users to review their responses and the AI's questions in detail.
- Session History: All interview sessions are saved, enabling users to track their progress and identify areas of improvement over time.
- Flexible Practice: Users can set custom session names and durations, making practice adaptable to their needs.
How we built it
PocketInterview is built as a native iOS application using SwiftUI for a modern and intuitive user interface. The core of our AI interview functionality is powered by the Tavus AI API, which provides the realistic conversational agents. For backend services, we utilize Supabase, handling user authentication, storing interview session data, transcripts, and AI-generated score details. The application integrates with the device's file system to allow for CV uploads, and a custom CV extractor processes these documents to inform the AI's personalized questioning. AI scoring for clarity, grammar, and substance is performed by an OpenRouter LLM (Deepseek R1) function deployed on Supabase Edge Functions, triggered by Tavus webhooks.
Challenges we ran into
Developing PocketInterview presented several challenges. Ensuring the accuracy and relevance of the AI feedback was paramount; this required extensive fine-tuning of prompts and careful integration with the LLM to provide genuinely constructive assessments. CV extraction and parsing proved complex due to the myriad of document formats and structures, necessitating robust text sanitization and pattern recognition. Integrating the Tavus API and managing its webhooks for real-time transcript and scoring updates also required meticulous handling of asynchronous operations and error states. Finally, optimizing the in-app WebView performance for the Tavus conversation interface was a continuous effort to ensure a smooth user experience.
Accomplishments that we're proud of
We are particularly proud of creating a highly personalized and realistic AI interview experience that genuinely adapts to the user's CV. The detailed and actionable feedback on clarity, grammar, and substance, powered by our custom AI scoring system, is a significant achievement, providing users with insights they can immediately apply. We are also proud of the seamless integration of complex AI and backend services into a user-friendly mobile application, making advanced interview preparation accessible to everyone. The ability to track progress through a comprehensive history and review full transcripts further enhances the learning journey.
What we learned
We learned the critical importance of contextual AI interaction in creating truly valuable learning experiences. Generic feedback is insufficient; personalization based on a user's background significantly enhances engagement and effectiveness. We also gained deep insights into the nuances of natural language processing for assessment, understanding that a multi-faceted scoring approach (clarity, grammar, substance) provides a more holistic and useful evaluation than a single overall score. Furthermore, the development process reinforced the value of robust backend infrastructure (Supabase) and flexible API integrations (Tavus, OpenRouter) for building scalable and feature-rich applications.
What's next for PocketInterview
For PocketInterview, our next steps include:
- Expanding AI Persona Library: Introducing more specialized AI interviewers for various industries and roles (e.g., Data Science, Marketing, Product Management).
- Advanced Feedback Mechanisms: Implementing features like sentiment analysis, pace detection, and non-verbal cue analysis (if technically feasible and privacy-compliant) to provide even richer feedback.
- Interview Coaching Modules: Developing guided learning paths and interactive modules based on common interview challenges.
- Multi-language Support: Extending the AI interview capabilities to support other languages, catering to a broader international audience.
- Web and Desktop Versions: Making PocketInterview accessible across more platforms to reach a wider user base.
Log in or sign up for Devpost to join the conversation.