Tassel

Ada marks up the student’s own paper: numbered steps trace an input, and a callout explains the loop that lets “1010” through.
Ada asks before she tells: she watches the student work, then nudges with a question instead of handing over the answer.
Ada watches the student’s handwritten DFA through Desk View and draws the correct machine, dead state included, on a shared whiteboard.
Describe your ideal tutor in your own words. Ada adapts her tone, pace and teaching style to match.
Ada steps an input through the machine one symbol at a time. Every answer if verified, so nothing is left to the model’s guess.
Every session is titled and tagged by subject automatically, so it’s easy to pick up where you left off.
Tassel works fully in English, Spanish and Korean, from the menus to the tutor’s voice.
Ada remembers how you learn across sessions, and you can read, edit or delete her notes anytime.
Ask for a diagram and Ada draws it live. Here she lays out three subnets behind their routers, with every IP range validated.
Customize your tutor’s name, hair, outfit, skin tone and voice. Every Tassel tutor is a woman, by design.

Inspiration

Stanford researchers found that girls taught by female math tutors showed significantly greater interest in STEM (https://nssa.stanford.edu/studies/impact-tutor-gender-match-girls-stem-interest-engagement-and-performance). Inspired by this, we built Tassel to make personalized mentorship more accessible through customizable AI tutors, multilingual support, live interactive learning, making STEM more welcoming & approachable for women of all backgrounds.

What it does

Tassel is a study companion that sees the work you're doing on paper, talks through problems with you, and brings explanations to life on a shared whiteboard. There are options for English, Spanish, and Korean, so you can learn in the language that feels most comfortable. With a customizable tutor (including the avatar, voice, and personality) that learns and remembers things about you, shows expressive features, and presented through a soft, colorful interface, Tassel makes studying feel more personal and welcoming.

How we built it

We built the desktop app with Electron, React, and TypeScript, backed by a Python FastAPI server and PostgreSQL. We connected voice (Deepgram -> speech to text -> 11ElevenLabs text to speech), vision to AI response so the tutor can follow a student’s work and explain it conversationally. The interface brings together a customizable tutor, camera input, and a shared whiteboard.

Challenges we ran into

getting the AI to accurately recognize handwritten work
unnatural speech patterns/conversations
tutor says they did something but they didn't
awkward pauses and general coordination
latency in responses
making the AI feel warmer

Accomplishments that we're proud of

We're proud of how welcoming Tassel feels. Other AI tutors have been done before, but our aesthetic gives Tassel a warmth and personality that make studying feel less intimidating. On the technical side, getting the tutor’s speech, avatar animations, and whiteboard drawings to stay in sync was a challenge, and we’re proud of bringing those pieces together.

What we learned

We learned that there's much more to tutoring than simply subject knowledge. Effective tutoring requires a deep understanding of the student's thought process and learning style. We also learned how challenging it is to connect image processing, natural conversation, and drawing in to a low latency experience.