AIgentum

Inspiration

Every semester, thousands of students face the same challenge:

• Endless hours scrolling through chair websites
• Multiple tabs open with scattered information
• Lost opportunities due to missed deadlines
• Overwhelming process of finding the right thesis

We asked: "What if AI could transform this weeks-long journey into seconds?"

What it does

AIgentum is your smart thesis assistant that simplifies the entire process. Here's how it works: Upload your CV and transcript, and our AI immediately understands your technical skills, academic background, and interests. Instead of you visiting countless chair websites, AIgentum searches through all TUM chairs and identifies the top 5 that perfectly match your profile. But it doesn't just list them, it provides you with a clear summary of what each chair expects, explains how your skills align with their requirements, and gives you direct links to their websites. To save you even more time and boost your chances of success, AIgentum goes a step further by providing personalized tips for your motivation letter and emails to supervisors, ensuring you make a strong first impression. What used to take weeks of searching, reading, and comparing is now transformed into a matter of minutes, helping you find not just any thesis position, but the one where you'll truly excel.

How we built it

We build a user-friendly interface using Python with Streamlit, making it intuitive and accessible for students. At its core, GPT-4 powers our intelligent agents, enabling natural conversations and deep understanding of student profiles and thesis requirements. We orchestrated these AI capabilities through LangChain, which helps manage the complex interactions between our different agents.

Challenges we ran into

One of our biggest challenges was dealing with the diverse and unstructured nature of chair websites – each with its unique layout and way of presenting information, making data extraction complex and demanding. We also grappled with ensuring our AI agents maintained context throughout conversations while delivering accurate, relevant matches.

There's a lot of potential for optimization to make AIgentum faster and more efficient, but we had to prioritize core features within our development timeline.

Accomplishments that we're proud of

Despite these challenges, we're particularly proud of what we've achieved. We successfully created an AI system that transforms a tedious, weeks-long thesis search process into a matter of minutes. Our multi-agent system not only understands student profiles with remarkable accuracy but also maintains context throughout the entire interaction, providing personalized and relevant recommendations

What we learned

We started thinking AI would do all the heavy lifting, but we quickly realized it's not just about cool technology - it's about really understanding what students go through when looking for a thesis. Working with AI tools like GPT-4 was like learning a new language; we had to figure out how to make it understand exactly what students need. It was pretty eye-opening to see how every chair has its own way of doing things. We learned that while AI is super helpful, adding a human touch to the process makes everything work so much better. It's like teaching a really smart friend to help other students find their perfect thesis match!

What's next for AIgentum

We're working on integrating AIgentum with TUM Moodle, allowing you to skip the manual upload of documents and just log in with your TUM ID. Once connected, AIgentum will automatically fetch your academic profile, courses, and grades, making your thesis search journey start with just a single click.

We're also exploring the addition of a supervisor interface, allowing professors to directly input thesis topics and requirements, creating a more dynamic and up-to-date database. Long-term, we aim including features like research paper recommendations and collaboration opportunities. Our ultimate goal is to make AIgentum the go-to platform for academic matching, transforming how Students and TUM connect for research opportunities.

Built With

• gpt-4
• langchain
• python
• streamlit