SkillMap

Inspiration

We noticed that when we first started learning how to code, we experienced a lot of difficulties since we didn't know where to start. Not that we couldn't find any resources, but because there was too many!! It got overwhelming and we didn't know where to start, how to progress or where to find actual good resources to get started. That's why we created SkillMap!

What it does

SkillMap is online web platform that allows users to put in what they want to learn, it can be anything from cooking to learning python, and SkillMap can generate a weekly plan for you to get started, along with resources and learning objectives! Once you finish your roadmap/plan, you can get certified as well! With a click of a button you are prompted to input details about your learning experience. If our AI detector detects an ingenuine response, then we won't give you the certificate! If you are geniune in your response, then we give you a certificate that rewards your hardwork!

How we built it

We built the frontend of our app using html, css and javascript. We have 3 main features in our backend:

Certificate Feature: We built this feature by using a premade image certificate, and then using Verbwire API to mint it as a omnichain NFT. We implemented exception handling, so if in a case where verbwire API doesn't work, we just default to using uuids. We implemented this feature so that each certificate issued can be unique as well as secure.
Proof of Learning Feature: But to get this certificate first, you need to write a response that is 350 characters long that will be checked by an AI that evaluates how genuine you are through langchain and openai technologies.
Roadmap Generator: Lastly, our main feature the roadmap generator is a RAG Agent(Retrieval-Augmented Generation) agent that leverages SerpAPI for web scraping similar topics entered by the user, FAISS for efficient vector similarity search, and OpenAI's language model to generate personalized study plans. It constructs a dynamic knowledge base by embedding web search results into a high-dimensional vector space, enabling rapid semantic retrieval of relevant information, which is then used to augment prompts fed into an LLM for contextualized content generation.

Challenges we ran into

Before implementing a RAG Agent, we initially thought of creating a fine tuned LLM using low rank adaptation. But since our resources were low, we couldn't proceed forward with this approach. It's also one of our team member's first hackathon, so they had to figure out a lot of frontend and implement it. We also had a dificult time integrating the frontend with the backend. Verbwire API also had several parts of its library on maintenance during this time, so that was a challenge as well.

Accomplishments that we're proud of

Sleek looking website UI
Nice animations
Created a fairly powerful RAG Agent
Integrating multiple features into 1 website
Had fun!

What we learned

Learned about NFT Minting
Learned about how to fine tune LLMs (even if we never used it)
Learned how to implement RAG Agents
Learned how to build a sleek website all in just pure html, css and javascript (not even bootstrap)