Try it out here!

About Surface Level

In surface level, we offer a survey that digs into you in reverse: Asking about your 'deeper' traits, and then relates your answers to real-world data to estimate your more 'surface level' traits, such as age, gender, and handedness.

Inspiration

Initially, the project started off as a one-off "unc detector" joke idea, where users could take a static quiz to receive an "unc score" or "unc meter", with a seperate ML component to determine survey questions and response weights in a colab notebook. Upon the theme reveal and establishing our team, we realized that some tweaks to the joke idea made a surprisingly decent plan that fit to our skill sets!

What it does

Users are presented with a quiz that measures your more abstract and deeper features (personality, behavior) and maps it out to more 'physical' ones, such as age, gender, and left/right handedness. If you ever wonder about your mental age, gender, among others, this quiz may be a fun exploration for you!

Our data is sourced from three datasets:

How we built it

Our project consists of two parts: A "personality quiz" with data visualization, and an ML component. We were fortunate that this resulted in everyone being able to pitch in a critical part of the project.

One of our team members learned HTML and CSS for the first time, and built the website layout by hand! We used a separate ML component on google colab to make JSON files to use as weights for our questionnaire component. Along with that, we implemented some graphs using d3 and svelte to display our data, as well as your place in it.

Some generative-ai assistance was leveraged during the research and Svelte coding.

Challenges we ran into

At first we wanted to incorporate unique user interactions into the age calculation, but no correlative studies existed in the field yet, so we decided this challenge was beyond our capabilities within the limited time frame, and chose to turn to demographic information.

Some technical challenges would be learning multiple new frameworks within the time frame, such as sveltekit, d3, csv-parser etc.

Accomplishments that we're proud of

Valerie had her first taste of web development at this hackathon, managing to learn HTML, CSS, and create the much-needed survey layout. She created a responsive HTML form that would become the questionnaire section of our project.

Sarah took on data visualization with svelte, arguably the trickiest feature we implemented. Using svelte, d3, and scrollytelling, she was able to utilize the csv files of the datasets we looked at, and display them in an interactive, engaging way with our survey results.

Lindsey took on the machine learning to narrow down questions for the survey, and assigning the proper weights/calculations for the survey results. To do this, she utilized linear and logistic regression in google colab, and exported JSON files to use as weights for the quiz. She also utilized svelte to make the quiz more interactive.

It was interesting to see the project unfold in a way that every member was able to contribute (and learn) a critical part for the project. From data visualization using Sveltekit, to basic ML, and web development fundamentals, we were all able to make a project we can all proudly say we did our part!

What we learned

Our team consisted of two members with knowledge in full-stack and ML, and one member who was beginning to learn how to make her very first website. Therefore, we all managed to take new knowledge home with us, ranging from learning Svelte, d3, (re-learning) ML techniques, and HTML + CSS.

What's next for Surface Level

Surface Level already expanded beyond what we initially planned, but if we had more time, we would have liked to integrate webcams and put a facial recognition feature (partly as a joke, but also due to interests!). Some more technical exploration on the ML side would have been interesting as well, as well as exploring more modern or appropriate datasets.

All in all, we are proud to have worked together, and to have deployed such a polished product for MadHacks 2026!

Built With

Share this project:

Updates

Submission history