Just Talk

Inspiration

We wanted to build a project centered around accessibility, particularly in situations where people experience stress, anxiety, or lack of confidence. Interviews are a universal experience, yet meaningful interview practice is often inaccessible. Many people are left practicing alone without clear feedback, which can make interviews feel intimidating rather than empowering.

As a team, we realized that we all needed interview practice ourselves, and we wanted a tool that could help users understand not just what they say, but how they say it. This inspired Just Talk — an interview practice platform designed to make feedback clear, visual, and accessible to everyone.

What it does

Just Talk helps users practice interview responses in a structured and reflective way. Users can enter an interview response, read it using a teleprompter-style interface, and record their video and audio simultaneously.

Once the recording is complete, the backend generates a transcript of the response and highlights filler words, pauses, and moments of hesitation. The video and transcript are displayed side-by-side, allowing users to click on highlighted sections and jump directly to those moments in the recording. This makes it easier for users to identify where they can improve and practice with intention.

How we built it

We intentionally chose a tech stack that we were curious about but not fully comfortable with, knowing it would challenge us and push us to learn.

Frontend: Next.js with TypeScript for a responsive and modern user experience
Backend: Spring Boot with Java to handle transcription processing and API logic
Database: PostgreSQL for storing transcripts and user session data
AI / Processing: Gemini API for speech-to-text transcription and analysis
Infrastructure: Docker and DigitalOcean for deployment and environment consistency

This architecture allowed us to separate concerns cleanly while experimenting with new technologies and deployment workflows.

Challenges we ran into

This project came with a steep learning curve. We experimented with and ultimately scrapped multiple approaches, including authentication solutions like Auth0, different transcription pipelines, and several DigitalOcean deployment configurations.

Working with unfamiliar technologies under hackathon time constraints was challenging. There were moments where features had to be simplified or removed so we could focus on delivering a working MVP. Debugging backend integrations, managing deployments, and aligning the frontend with backend APIs required constant iteration and teamwork.

Accomplishments that we're proud of

Building a fully functional MVP under time pressure
Successfully integrating video recording, transcription, and analysis
Creating a clean, intuitive interface that emphasizes accessibility
Tackling a new tech stack and learning it together as a team
Turning a personal pain point into a meaningful product

What we learned

One of the most important outcomes of this project was how intentionally we came together as a team to learn unfamiliar technologies. Rather than choosing tools we already knew, we decided to challenge ourselves by exploring a stack we had always been curious about.

This decision came with setbacks — features were scrapped, approaches were rethought, and progress sometimes felt slow. However, we stayed determined and committed to understanding the technology instead of taking shortcuts. We supported one another through debugging, research, and experimentation.

Through this process, we learned not only new frameworks and tools, but also how to collaborate effectively under uncertainty. We learned how to communicate clearly when stuck, break down complex problems, and remain adaptable when plans changed. Most importantly, we gained confidence in our ability to learn and grow together as a team.