Rowan

Inspiration

Interviewing can be stressful, especially when you are preparing on your own without much guidance. We wanted to make interview prep feel more accessible, personalized, and less intimidating by building a tool that helps people practice with confidence for a specific role.

What it does

Rowan is an AI-powered interview prep platform. Users upload their resume and provide details about a job, including the company, role title, posting link, and description. Rowan then crawls the web for relevant company and role information to create a more tailored interview prep experience.

It supports interactive mock interviews and gives feedback not just on what you say, but also on how you say it. Rowan combines role-specific research, live practice, and body language analysis to help users prepare in a way that feels much closer to a real interview.

How we built it

We built Rowan by combining several tools into one pipeline. We used Browser Use for web crawling and company research, OpenAI’s realtime API for the live interview experience, Supabase for the backend and authentication, and Twelve Labs for video analysis and body language feedback.

A big part of the project was connecting these systems into a smooth workflow, from resume and job input to research, mock interview, feedback, and iteration.

Challenges we ran into

One major challenge was integrating several APIs into a single experience that felt seamless and responsive. Setting up the realtime interview flow required careful handling of authentication, state, and latency.

We also spent time refining the Browser Use agent so it could consistently return high-quality, relevant information, and figuring out how to configure and prompt Twelve Labs to generate feedback that was actually useful.

Accomplishments that we're proud of

We are proud that Rowan is more than a rough prototype. It is a fairly involved application that combines web crawling, live voice interaction, video processing, and personalized feedback into one cohesive experience. We are especially proud of the pipelines we built to make these systems work together reliably in a way that feels useful to the user.

What we learned

We learned that building a strong AI product is not just about using powerful models. A lot of the work is in orchestration, prompt design, and making the output genuinely actionable. We also learned that personalization matters a lot. The better the system understands the candidate, company, and role, the more valuable the interview prep becomes.

What's next for Rowan

Next, we want to improve the quality of the feedback, especially around body language and delivery. We also want to make the mock interview experience more realistic and interactive, for example by giving the interviewer a stronger visual presence. Longer term, we want Rowan to feel even more like a realistic, adaptive interview coach.

Built With

• browseruse
• openai