Inspiration

Solving problems inside the MLOps domain is really hard; Tooling and problem solving skills required for the following things -> Docker, bash scripting and AWS, FastAPI

Christos and Mert were taking the course from https://mlops-club.org and said why don't we have an assistant that retrieves info from the courses.

What it does

Currently, it just takes one question and answer it with a retrieved information from the pinecone index that populated the course scripts!

Challenges we ran into

arize phoenix context wasn't resetting: it just doesn't finish the previous ctx, that's why it was just tracing the object with the old-project's context

Accomplishments that we're proud of

Created a Video Transcript Scraper thats prrd like a kitten.

Pinecone VectorStore script was so modular, even though we had spend a lot of our time on the data cleaning + creating the VectorDB; it was really good

We have able to run Llama Workflow to perform an Agentic RAG.

We were able to show a Proof-of-Concept to share the Best Practices for Software Engineering that we all need, especially during a hackathon 🥵

What we learned

  • Llama Workflows
  • Arize Phoenix tracing
  • Llama Deploy

What's next for mlops-club-rag

  • Create UI
  • Create Discord Bot
  • Connect to more mlops course content, include high quality metadata and links to direct users to continue learning!

Built With

  • fastapi
  • llamaindex
  • pinecone
Share this project:

Updates