Sustainify

Customizing the importance of each factor as per user preference
Green Score computed for the current product
In-depth explanation for a specific category score
Stash - dashboard to see past purchases and your own tree powered by your overall sustainability score
Dark mode and subscores for categories of products
Share your Sustainify Score and plant to friends and family
Sustainify Plant

Inspiration

As the e-commerce industry expands in all business districts, buying endless amounts of products has never been easier. Shoppers are heavily prioritizing quick, temporary solutions rather than doing the research for a more sustainable option.

What it does

When shopping online, Sustainify's browser extension effortlessly scans the product and gives a sustainability score based on material composition, transport emissions, ethical sourcing, community approval, among other factors inspired by the commonly used ESG framework. These scores are accompanied by a brief justification and displayed to the shopper in an intuitive manner. Additionally, Sustainify searches the web and provides product links to more sustainable alternative products to consider instead of the current one.

Beyond providing sustainability scores to products, Sustainify also allows shoppers to keep track of their online purchases through their own personalized Stash. If shoppers choose to buy the product, they can save it to their Stash which will immediately update the users personal Stash and Sustainify Score on their dashboard. Here the users online purchases are categorized and given a subscore contributing to their overall Sustainify Score. This allows the user to improve their sustainability on specific areas of their online purchases. Each product in the Stash can be is linked to the original product page which enables users to quickly access more information.

Finally, users have their own personal Sustainify plant that is linked to their Sustainify Score. Depending on the users score, their plant's health and growth will reflect how sustainable the user's online purchases are. This adds a gamified experience that allows users to share and encourage their friends to shop more sustainably.

How we built it

Our frontend is framework-less and entirely made from scratch using JavaScript, HTML, and CSS. The backend logic is powered by three work flows made using Gumloop. These work flows detect product pages, calculate sustainability scores, and recommend alternative sustainable products based on extracted data.

Challenges we ran into

Having abundant access to LLMs and AI agents, we wanted to optimize their usage in the application to be more sustainable and also make the user experience faster. This involved continuous reviews of our AI workflows and using programming logic when possible instead of delegating basic tasks to LLMs.

For our product recommendation feature, striking a balance between in-depth web search to surface level suggestions was perhaps the most challenging technical decision. Searching the entire web to get good suggestions took too long and also did not fit into our core sustainability objective. However, creating a manual database with sustainable products was unrealistic for the horizontal scale of products we wanted to apply Sustainify to. Hence, we decided on temporarily using Reddit as our main source for alternatives. By searching through subreddits related to sustainable products, we were able to provide a wide range of alternatives for popular products.

Accomplishments that we're proud of

In a world where AI at the forefront of software development and business, sustainability is something our team believes needs to be considered more before it is too late. We're proud to make an extension that encourages people to take sustainability into account when shopping online.

On the technical side:

creating an intuitive beautiful UI with a light and dark mode that has potential for more "gamification." This involved creating an ecosystem of features: the dropdown when you click on the extension (for settings), the toast notifying users that a page is being analyzed by AI, the injected analysis panel, and the dashboard personal homepage. We used chrome.storage to handle data persistence, utilizing syncing so that changing theme in one of the previously mentioned locations instantly updates it in any of the others, and similarly have instantaneous dashboard reactivity to new items added to stash.
our extension is framework-less and made in-house. This made it more difficult to build, but the reward is a light and modular design that is easy to move around. Our frontend is powerful rather than relying on backend calls; for example, we used a functional JS DFS pattern to scrape all text out of the page's source to make AI calls much lighter.
synchronizing different AI workflows seamlessly into a single product took multiple iterations and cautious design choices. We save time by concurrently calling multiple workflows at once, creating a surprisingly fast timeline for the user to see detailed breakdowns of the product's background checks.
recommending products with a higher sustainability score is a feature that could be strengthened, as link hallucinations is a problem. We limited the AI to just reddit domains in order to get the most relevant results.

What we learned

We learned the ins and outs of making a web extension, web scraping, using browser state to store information and developing complex AI workflows to automate the manual process of scoring products on a comprehensive set of factors.

What's next for Sustainify

The core next step for Sustainify is to find a concrete and realistic method of giving sustainability scores to every product online. Our research during the hackathon led us to explore ESG, combining it with our own knowledge of sustainability, we were able to create a reasonable estimate for sustainability of products. However, we acknowledge that their is no definitive way to quantify this. Hence, we would love to do further research and perhaps talk to industry experts about our options to improve this metric.