Hypernym | Devpost

see more in the google slides link pasted below if you're curious :)

Inspiration

We write a lot of outbound messages at Hypernym as a new startup. So, using the sponsor's technology, we focused on how we can utilize their tools in an agentic workflow for writing content/outbound emails, titles, agents to have understanding of yours and competitors journeys to make informed decisions and actions by itself. This is something we can use ourselves in the future!

What it does

tell if agents or humans need to understand the enriched data better. Be more human or agentic depending on the purpose, given the datastream--> what should be said differently. Compare different messaging threads in order to determine what drift is occurring and how to bring the LLM back into orbit.

How we built it

We built end to end workflow using APIs from Mixpanel, Bright Data, Arcade, and Hypernym's own API.

Challenges we ran into

The dummy data provided was not very substantial, and even having more data sources would have greatly improved our results to demo.

Accomplishments that we're proud of

A half technical, half non-technical team, we managed to finish the demo on time. Really excited

What we learned

Less time planning, more time executing, and managing all the tasks effectively as a team with clear communication could've saved us from finishing the development at the end slowly.