AceSAT: Adaptive SAT Prep & Personalized AI Study Coach

A Socratic AI tutoring agent that diagnoses student weaknesses, creates custom weekly plans, and provides progressive scaffolding to close the public school tutoring gap.

Comment

Inspiration

Wealthy families spend thousands of dollars on private SAT tutors and structured college prep courses. For students in underserved public schools, these resources are entirely out of reach. Standard practice websites often provide direct answers immediately, which encourages passive copy-pasting, or they present flat questions that easily demotivate students.

We were inspired to build AceSAT to democratize elite academic preparation. By structuring the application as an active, state-driven pedagogical agent, we created a tireless, private AI-tutor that adjusts to each student's pace and guides them through concept mastery using progressive Socratic scaffolding.

What it does

AceSAT acts as an interactive learning companion split into four integrated modules:

  1. Coach Onboarding: Students converse with Coach Ace to establish target scores and weekly study plan blocks.
  2. Adaptive Prep Workspace: The agent dynamically serves practice questions. Correct answers boost mastery and escalate question difficulty (Easy ➔ Medium ➔ Hard). Incorrect answers lower the difficulty to reinforce fundamentals.
  3. Scaffolding Engine: When a student gets stuck, they can request step-by-step hints. Instead of giving away the answer, the agent guides them through:
    • Level 1: A conceptual formula or grammar reminder.
    • Level 2: A setup helper guiding the first step.
    • Level 3: A near-solution calculation check.
  4. Interactive Knowledge Graph: Renders a coordinate-based SVG map of SAT prerequisites, color-coding mastery levels (Red/Yellow/Green) in real-time.
  5. Agent Decision Console: Displays a live telemetry terminal showing the agent's calculations, difficulty routing, and diagnostics behind the scenes.

How we built it

We built AceSAT as a lightweight, fluid Single Page Application (SPA) to ensure compatibility with older Chromebooks and mobile connections:

  • Frontend: React + Vite + Vanilla CSS.
  • Design & Animations: Frosted-glass navigation headers, staggered card transitions, and a spring-based shake animation ($\text{cubic-bezier}$) that vibrates incorrect options to provide tactile visual feedback.
  • AI Core: Integrates Google AI Studio's stable API (gemini-flash-latest) using client-side environment configurations.

For example, when working on circle equations in the geometry topic, the agent guides the student through completing the square to find: $$(x - h)^2 + (y - k)^2 = r^2$$

Challenges we faced

Migrating between restricted model endpoints on newer Google API accounts was a key hurdle. We resolved this by building a dynamic diagnostics script to poll available endpoints, switching our core engine to the stable gemini-flash-latest pointer. Additionally, designing the agent console required creating custom event emitter logging hooks to capture client-side state transitions in real time.

What we learned

We learned the value of "Mastery Learning" and scaffolding over direct answers. We also learned how to build high-performance client-side state machines to run diagnostic routing trees without needing heavy database backends.

What's next for AceSAT

We plan to expand AceSAT beyond SAT preparation to include foundational AP Science and AP History curriculums, introduce multilingual Socratic support for ESL students, and build classroom dashboards so teachers in public schools can track group competency paths in real time.

Built With

Share this project:

Updates

Submission history