Pathfinder

✨ Inspiration

Pathfinder was inspired by the concept of "Advisor at Your Hands"—a tool aimed at providing accurate and reliable career guidance without compromising the quality of information. Our goal was to create an AI chatbot that helps users make informed career choices while promoting the unique programs offered by Penn State Smeal College of Business 📚. Our top priority was ensuring that Pathfinder avoids "hallucination"—the generation of misleading or incorrect information—by strictly relying on user-provided data and its programmed knowledge base.

The RAG bot for the Institute for Computational and Data Sciences (ICDS) 🧠 was developed out of the need for a tool that provides precise answers without fabricating information, particularly when summarizing complex documents. This feature ensures users receive trustworthy and well-informed responses.

🤖 What it does

Pathfinder is an AI-powered platform with two key functionalities. Career Guidance Chatbot: The AI career advisor helps users explore different career paths, provides personalized advice, and offers insights tailored to individual interests and skills 🎓. It also promotes Penn State Smeal programs, ensuring users are aware of educational opportunities that can enhance their career prospects.

RAG Bot for ICDS: This component employs Retrieval-Augmented Generation to summarize and provide accurate responses based on user guides or documentation 📑. It synthesizes information from multiple sources, offering the most relevant and concise answers without losing essential context.

🛠️ How we built it

Pathfinder was built using Next.js for the user interface, enabling a dynamic, interactive, and user-friendly experience 🌐. We utilized Langchain for the AI backend to integrate sophisticated natural language processing and RAG capabilities.

To achieve the chatbot’s dual functionality, we established two distinct data pipelines. Career Advisor Pipeline: This pipeline pulls data from curated sources on the Internet and Smeal program details, ensuring that the chatbot provides users with well-rounded and accurate career guidance 🎯. RAG for ICDS Pipeline: We implemented a multi-step process where the bot first retrieves relevant sections from user guides, performs detailed summarizations, and presents synthesized information to the user for maximum clarity and precision 📝.

🚧 Challenges we ran into

The development journey came with its fair share of challenges.

Getting Started with Langchain: We encountered difficulties initially, as nothing seemed to work seamlessly. We tested over four models, trying out different implementations to find the best fit for our purposes. Integrating RAG for Summarization: Summarizing complex documents while maintaining context and detail was technically demanding. We fine-tuned the bot’s algorithms to ensure reliable and meaningful outputs 📜, while also managing token limitations to prevent errors. Linking Front End and Backend: Integrating the front end with the back end proved challenging due to deprecated libraries and inconsistencies between different versions of functions in Next.js 🔄.

🏆 Accomplishments that we're proud of

We successfully developed a dual-purpose AI chatbot that sets a high standard for accuracy and reliability. Pathfinder serves as both a dependable career guide and a robust tool for summarizing complex information for ICDS, making data more accessible and easier to understand 🌟. The RAG bot’s ability to condense multiple document readings into one cohesive response was a significant technical achievement, and we’re proud of the positive feedback from initial testers who praised its user-friendliness and value 🙌.

📚 What we learned

Building Pathfinder provided valuable lessons about developing AI systems focused on data integrity and user trust 🤝. We gained deep insights into implementing RAG techniques, particularly in combining information retrieval with natural language summarization. Working with Next.js and Langchain broadened our knowledge of creating scalable AI-driven web applications 💻. User feedback played a crucial role in refining Pathfinder, highlighting the importance of an iterative, user-centered development approach.

🚀 What's next for Pathfinder

Looking ahead, we plan to expand Pathfinder’s capabilities by incorporating additional data sources for career guidance and enhancing the RAG bot’s summarization techniques to handle a wider variety of documents 🔍. We also aim to introduce real-time updates to keep the knowledge base current and explore machine learning models for even more personalized recommendations 📈. Ultimately, our vision is for Pathfinder to be a trusted resource for career advice and data insights, continually supporting the educational missions of Penn State Smeal and ICDS 🎓. We also plan to add a web scraper for improved user convenience and seamless data integration.

Built With

• ai
• css
• html5
• javascript
• langchain
• llm
• next.js
• node.js
• npm
• openai
• react
• tailwind
• typescript
• vercel
• zod