-
-
Drop your transcript. That's it. CometAdvisor instantly knows your entire degree history, GPA, and what you still need to graduate.
-
Your AI advisor greets you by name, knows your progress, and starts the conversation no appointment needed, no hold on your account.
-
A full personalized semester plan built from real UTD course data, not generic advice your specific path to graduation.
Inspiration
Every UTD student knows the feeling. Registration opens in two weeks, you have an academic hold on your account, and your advisor is booked solid for the next 4-5 weeks. Your options? Hope for a walk-in slot that may not exist, or send an email just to check a box leaving you with a confusing 4-page Excel degree plan you half understand.
We built CometAdvisor because we lived this problem. Students shouldn't need to fight for 30 minutes with an advisor just to know what classes to take next semester.
What It Does
CometAdvisor is an AI-powered voice academic advisor for UTD students. You upload your unofficial transcript, speak to the advisor naturally, and it generates a fully personalized semester plan what to take next, which professor to take it with, and exactly why.
- Voice-first interaction powered by ElevenLabs
- Transcript parsing upload your PDF, we handle the rest
- Visual degree roadmap see your
How We Built It
Comet Advisor is built on a Next.js and Tailwind CSS frontend with a Python FastAPI backend. Students upload their UTD transcript which gets parsed using a custom PDF parser that extracts completed course codes. We use the Nebula Labs API to pull real UTD course data and professor information, and Google Gemini as our AI brain to interpret the student's goals and generate personalized course recommendations grounded in real degree flowcharts we hardcoded for each major. ElevenLabs powers Cali, our voice advisor, creating a natural turn-based conversation. The entire plan builds in real time on a drag-and-drop dashboard as Cali speaks.
Challenges We Ran Into
Our biggest challenge was making the voice conversation feel natural. The Web Speech API has built-in timeouts and would cut off mid-sentence or activate while Cali was still speaking, we had to build a custom silence detection system and strict turn-based guards to prevent the mic from bleeding into Cali's responses. We also ran into merge conflicts coordinating between three separate branches, frontend, backend, and transcript parser all being built simultaneously under time pressure. Finally, bridging Cali's natural language responses to structured course data was tricky — we had to build a regex parser that extracts course codes from conversational text and maps them to the right semester columns in real time.
Accomplishments We're Proud Of
We're proud that we shipped a fully functional voice-first AI advisor in under 48 hours. The integration between ElevenLabs, Gemini, and the Nebula Labs API creates something that genuinely feels like talking to a knowledgeable advisor, not a chatbot. We're also proud of the real-time course plan that builds as Cali speaks watching courses animate into the board during a live conversation is a genuinely magical demo moment. And the UI built to match the Nebula Labs design language, looks like a real product, not a hackathon project.
What We Learned
We learned that voice UX is significantly harder than text UX, the margin for error is much smaller and latency issues that would be invisible in a text app become obvious and jarring in a voice conversation. We also learned the value of locking down a shared architecture early, our biggest time losses came from parallel frontend and backend development that diverged before we had agreed on API contracts. On the technical side, we got deep experience with the Web Speech API, ElevenLabs streaming, and Next.js API routes proxying to a Python FastAPI backend.
What's Next for Comet Advisor
The immediate next step is expanding our degree flowchart database to cover every UTD major and minor, right now we have CS and SE hardcoded, but the architecture supports any major. We also want to integrate the Nebula Labs professor rating data so Cali can recommend not just the right courses but the right sections with the best professors. Longer term, we want to add a scheduling layer that cross-references the Nebula Rooms API so Cali can recommend courses that actually fit together in a student's weekly schedule. The dream is a single conversation that takes a student from confused freshman to fully planned four-year roadmap in under five minutes.
Log in or sign up for Devpost to join the conversation.