RAGTag

Inspiration

Starting a business is stressful, a lot of needs need to be met, and a balance needs to be struck. Being presented with mountains of LinkedIn profiles can be overwhelming, that's where RAGTag comes in.

RagTag means messy and disorganized, but our solution is anything but (it's ironic!). Building a team can be stressful, and there are so many things you need to consider. We tend to pick people we know or get along with instead of those who truly suit our team needs. That's why RagTag was created.

What it does

RAGTag takes in your LinkedIn profile, and automatically finds your perfect teammate on LinkedIn. Someone who complements your weaknesses, your other half <3. Using RAG, we parse LinkedIn profiles, combined with generative capabilities, we find your perfect teammates, as well as why.

How we built it

First we generated a bunch of LinkedIn profiles, and embedded them.

The user then sends in their profile, and we use a pipeline to convert it to a string, and then summarize it using Coheres summary endpoint.

That summary is then passed to a query, which queries the vector database to find profiles with complimentary skillsets, along with the Cohere rerank api.

Coheres command LLM is then used to interpret the search results, and return the top 3 candidates, swell an explanation of why they compliment your skill set.

Challenges we ran into

How are we supposed to query the database when we don't want similar teamates but rather different teammates???

Tried using cosine distance instead of cosine similarity, different MMR metrics. What ended up working the best was passing in a summary of the users profile, and writing "not that".

How do we justify our decisions??

We ask the chatbot to tell us why they chose those candidates, and list the reasons.

Accomplishments that we're proud of

We built a working application in less then 3 hours. We didn't do a traditional document search, but rather something a bit more off the beaten path.

What we learned

What's RAG, and how to use Cohere's embed and re-rank API to implement RAG
Cohere's embed, re-rank and summarize API
Finding a perfect teammate is really complicated, and in reality there is no right answer to what makes the perfect teammate.