Before UC Davis I attended a community college in San Francisco. Deciding which university to attend and which major to pursue, I felt overwhelmed by the number of options and by the articulation agreements between colleges and universities.

With so many students preparing to transfer, my counselors did not always have time to meet with me, and our appointments were short. I started using ChatGPT as an extension of my advising, to filter information about prerequisites, courses and grades and to reach the right links faster. It helped, but I kept thinking that I wished I could standardize those conversations, so that everyone going through the same transition could start from a conversation already full of context.

My experience was not unusual. The American School Counselor Association has recommended one counselor for every 250 students since 1965. In the 2023/24 school year the national average was 376:1, with 49,404,386 public-school students sharing 131,230 counselors, and that average hides a massive spread between states.

Caseloads are only half the story. Even when a counselor exists, college advising competes with schedules, testing, discipline and mental-health crises, and private schools invest in it far more than public ones.

Only 3 of 51 (50 states and D.C.) meet the 250:1 standard. The map shows every state's students-per-counselor ratio, with the worst, the best and California called out.

Numbers show the gap, and students explain what it feels like. Before designing anything I went looking for their side of the story online, in forums where students talk about applying and transferring.

I also talked with local students at City College of San Francisco who were going through the same transfer process I had, and collected more answers through a Google Forms survey. These are the themes that kept coming up.

The idea stayed an idea until my Design Drawing class at UC Davis in Winter 2026, whose final project let us sketch a UI/UX experience. Seeing how quickly good ideas came out of rapid sketches, I started drawing the advisor I had always wanted.

Without a whiteboard, I used the blank sheets left over from class, drew the screens one by one and taped them to the wall of my dorm room, so I could see the whole experience and the flow between screens at once.

Each path asks 12 to 13 questions. These are the screens as they moved from paper to Figma to the shipped product, where every step gained its own Obee illustration and a free-text field that accepts anything the buttons can't capture.

Generic chatbots answer confidently even when they are guessing. Obee does the opposite. Each text box calls Claude to extract profile data from the intake and return a structured recommendation grounded in the documentation already in Obee's brain, including IGETC, TAG, ASSIST.org and Transferology.com, which is all the context I once wished I could bring into my own AI conversations.

If information is missing, Obee asks a question instead of making something up. Once the profile is complete it classifies schools, explains its reasoning and suggests next steps for students who want to strengthen their profile.

Every intake costs money in API calls, and I cover the infrastructure myself so Obee can stay free for students. It is my way of giving others the tool I wish I had when I was transferring.

Growth has been entirely organic, to the point that Obee reached the first page of Google for "Obee AI". The product is complete, has regular users, and the API budget I set aside keeps it running for students.