BrandCast

Inspiration

We felt there was a gap in the advertising market when it came to streamers. Far too often while watching videos, we are hit by irrelevant adverts that disinterest us and make us turn away. While no-one wants adverts, we thought the ones there should be as best tailored to streamer as possible by big brands as a traditional to static advertising.

What it does

Simplifies a previously very arduous process of looking manually at loads of people individually to identify potential and existing affiliations to the click of a button.

How we built it

Built using Twitch web scraping and API access to gather data on streamers. Sentiment of streamers and their compatibility with a client advertiser is analysed by Claude LLM. Front-end built using Next-JS and Chakra UI. Also used a language vector embeddings model to enable the use of a semantic search to successfully match possible streamers to the client company's profile.

Challenges we ran into

Front-end api calling difficulties. Some issues with the api and cleansing data. Some issues with semantic search. All eventually resolved :)

Accomplishments that we're proud of

Fixing the bugs and making a unique product! A guy at theTradeDesk said he hadn't heard of anything like it before! We feel there are really good commercial opportunities here.

What we learned

Tonnes about JS and the streaming/advertising industry.

What's next for BrandCast

Let's see!

Built With

• claude
• express.js
• nextjs