theinterbuddy

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for theinterbuddy

Inspiration

Interviews are one of the most important milestones in anyone’s career. They decide whether a person gets their dream job, a meaningful internship, or even the first role that will define their career path. Yet interviews are also one of the most stressful experiences. Many candidates struggle with the pressure, the unpredictability of questions, and the lack of structured feedback. Preparing for interviews often means flipping through long lists of practice questions, watching videos, or rehearsing in front of a mirror. These methods help a little, but they do not create the same environment as a real interview. They lack the back and forth interaction, the realistic time pressure, and most importantly the constructive feedback that helps candidates improve.

The inspiration for our project came directly from our own journeys as students and early professionals. We often found ourselves sitting late at night before an interview, wishing there was someone who could act like a mock interviewer on demand. Friends were busy, mentors were not always available, and professional coaching services were either expensive or inaccessible. At the same time, we realized that modern artificial intelligence is powerful enough to replicate the role of an interviewer. This gave us the spark: what if we could build a personal “interview buddy” that anyone could access instantly, that behaves like a real interviewer, asks thoughtful questions, and provides actionable feedback?

That idea became the foundation of theinterbuddy.

What it does

Theinterbuddy is an AI powered interview coach that combines behavioral and technical practice in one seamless platform. A user begins by selecting their target company, role, and difficulty level. Based on this setup, the system generates interview questions that are personalized to the chosen role. For behavioral practice, the AI avatar asks questions just like a hiring manager would, focusing on past experiences, collaboration, and problem solving skills. For technical practice, it introduces role specific challenges and even coding questions.

The most important part is the feedback loop. After each answer, whether spoken or coded, the system generates constructive feedback. For behavioral responses, it analyzes clarity, structure, and relevance. For technical or coding responses, it evaluates correctness, time and space complexity, and code style. It does not stop at criticism but provides specific strengths, areas of improvement, and even corrected code if needed. To make the practice feel immersive, we added AI generated video avatars that ask the questions and deliver the feedback, simulating the flow of a real video interview.

How we built it

We built theinterbuddy using Flask for the backend and SQLAlchemy for database management. The application manages user accounts, interview setups, and progress tracking. We integrated OpenAI models for generating questions, analyzing answers, and producing structured feedback. For coding questions and evaluations, we designed prompts that request feedback in JSON format, which makes it easier to display results in a clean and actionable way.

To add realism, we integrated the D-ID API to generate talking avatar videos. This allows the interviewer to appear as a real human face delivering questions and feedback. We handled transcription of user answers with the Whisper model, ensuring that spoken answers could be converted into text for feedback analysis. The frontend was built with a focus on design and usability. We used modern responsive layouts, animated gradients, and particle backgrounds to make the platform feel polished and engaging.

Challenges we ran into

One of the biggest challenges was ensuring reliability in feedback generation. Large language models can sometimes be inconsistent. We solved this by structuring prompts carefully, enforcing JSON outputs, and adding fallback logic for both questions and feedback. Another challenge was integrating audio and video processing seamlessly. Handling audio uploads, file size checks, and ensuring smooth transcriptions required careful error handling. Video generation also needed polling and graceful fallbacks when the external API was slow or unavailable.

We also faced challenges in designing a user experience that felt supportive rather than intimidating. Our goal was not to overwhelm users with harsh criticism but to balance encouragement with actionable advice. Achieving this tone consistently required tuning prompts and adjusting the way we displayed results.

Accomplishments we are proud of

We are proud that we were able to combine multiple advanced technologies into one cohesive platform within the hackathon timeline. We managed to create a system that not only asks questions but also adapts to user input, gives feedback, and visualizes it through realistic avatars. Seeing our project come together into something that feels like a real interview coach is deeply rewarding.

What we learned

Through building this project we learned how to integrate several APIs into one functioning product. We strengthened our understanding of Flask and SQLAlchemy, improved our skills with frontend design, and became better at engineering prompts for AI models. We also learned how important it is to think about the user journey from start to finish. Interview preparation is already stressful, so building an interface that feels safe, supportive, and motivating is just as important as the underlying technology.

What is next

Looking ahead, we plan to extend theinterbuddy by adding more role specific question sets, support for additional programming languages, and advanced analytics dashboards for tracking long term progress. We also want to experiment with more expressive avatars and voice customization to make the experience even closer to a real interview. In the future, we see this project becoming a go to platform for students, job seekers, and professionals who want to practice and improve continuously.

Built With

• did
• flask
• html/css/javascript
• openai
• sqlalchemy
• whisper