tuki | Devpost

Inspiration

Have you ever had to look for sponsors for an event? Have you ever had to cold call companies asking for partnerships? Are you looking to promote your products at events? Don't know who has the same company values and would work best with you?

What it does

Tuki is an interactive web app with an effortlessly navigatable system and a unique matching algorithm that makes building connections easier than ever.

How we built it

Based on a company and organization's mission statements, values, and goals, Tuki's matching algorithm will measure similarity using a similarity score. We measured similarity scores by cosine similarity and BERT token embeddings to detect similarities in semantics. Considering mission statements, values, and goals with a specific weighting system, our matching algorithm ensures quality long-term connections between organizations and companies. Our matching algorithm and other backend components were then prototyped in an interactive window.

We used Figma to prototype website design elements and interactions. Our prototype demonstrates how organizations can request a connection with companies, and how companies are able to customize information on their available resources/sponsors. Additionally, we used Figma to model how organizations can post their event, allowing companies to connect with specific initiatives.

Lastly, we used HTML and CSS to create the front-end portion of our website based on our Figma prototype.

Challenges we ran into

Planning website structure: While planning our prototype, we had a difficult time organizing and incorporating all desired features into one cohesive product, while also aiming for easy navigation for users. We created multiple drafts of what certain pages would look like, as well as how we wanted to structure the pages.

Backend: Initially, we focused our matching algorithm on the structure of words/sentence structure; however, we realized that was not our intent to match free response content, such as mission statements. We extensively researched ways to capture similarities in semantics and decided to use BERT with tokenized word embeddings. While we tried to use SBERT, a significantly improved framework, it could not be installed on our computer.

Accomplishments that we're proud of

-Successfully creating a matching algorithm that detects semantics -Animations in Figma + cohesive theme of our product -Successfully integrated many parts learned from classes/workshops into our project -Alex programmed the longest piece of runnable code he's ever made thus far!!

What we learned

We learned that having a clear framework and timeline is crucial to the development of a project, especially in a short timeframe. Before splitting into our respective roles, we clearly laid out how we envisioned the product to work, putting ourselves into the shoes of users in our targetted audience.

We realized the importance of patience and communication; when certain things did not go as planned, we persisted in solving the problem, communicating to each other potential updates to the timeline. When we were eager to build additional features that strayed away from the main goal, we were patient, noting them down for later and prioritizing aspects essential to the project.

What's next for tuki

-Conduct interviews/surveys with our target audience to gain feedback on current designs/components -Add messaging feature and calling/interview feature to create more accessible connection between companies and organizations -Optimize matching software -Connect backend with front end