Frizzle

Frizzle is an AI-powered platform that functions as an operating system for K-12 math classrooms. It leverages computer vision and large language models (LLMs) to read and analyze handwritten student work from paper with a documented 97% accuracy rate. The product is designed for math teachers, instructional coaches, and school or district administrators who need granular, standards-level formative assessment data without disrupting existing paper-based workflows. Students continue writing on paper with pencils, and teachers simply photograph stacks of work using a phone, document camera, or scanner. Within approximately eight minutes per class, Frizzle processes every page, links each to the correct student, and generates live dashboards showing which Common Core State Standards (CCSS) each student and class has actually mastered. The system also includes a confidence-interval mechanism that flags uncertain grades for human review, ensuring reliability. Frizzle is currently operational in over 30 schools and districts, including a college math pilot at Vanderbilt University and Arizona State University. The platform is FERPA and COPPA compliant, with SOC 2 Type II annual audits, and has processed over 100,000 questions to date. Its core value proposition is reclaiming the 10-15 hours per week teachers spend grading by hand while providing real-time, actionable data on student thinking, misconceptions, and prerequisite gaps.

Computer Vision Handwriting Recognition

Frizzle's computer vision system is trained on a dataset of 1.4 million pages of K-12 student work. It recognizes print, cursive, scribbled, and sideways handwriting without requiring students to change their writing style. The system parses each step of student work, not just the final answer, and understands multiple valid solution paths. For example, a quadratic equation can be solved via factoring, square roots, or the quadratic formula, and Frizzle credits all three approaches correctly. The accuracy rate for grade assignment is 97%, with a built-in confidence interval that flags ambiguous responses for manual teacher review.

Step-Level Misconception Detection

Unlike basic answer-checking tools, Frizzle analyzes the entire solution process to identify where thinking went wrong. The system has 147 named misconceptions across K-12 mathematics, each mapped to specific CCSS or TEKS standards. When a student makes an error, Frizzle tags the exact stroke on the page and links it to the underlying misconception, such as sign errors, distributive property mistakes, or prerequisite gaps from earlier grade levels. This feature enables teachers to provide targeted feedback rather than generic "wrong" marks.

Live Standards-Level Analytics Dashboard

Frizzle generates live dashboards that aggregate student performance data across a single class period or an entire district. The dashboards display mastery percentages for each standard, categorize students into mastered, developing, and at-risk groups, and track engagement metrics. Coaches and administrators can see which misconceptions are spreading across classrooms, identify equity gaps in real time, and make data-driven decisions about curriculum adjustments. The system supports CCSS, TEKS, and over 30 additional state frameworks, and is curriculum-agnostic, working with Eureka, Illustrative Mathematics, Saxon, and other programs.

Automated Student Work Linking and Ingestion

When a teacher snaps a stack of papers using a phone, document camera, or scanner, Frizzle automatically links every page to the correct student profile within approximately 30 seconds. The system ingests classroom photos and associates each submission with the appropriate student, eliminating manual sorting and data entry. This feature requires no student logins, no tablets, and no migration to digital workflows. Teachers simply photograph the existing paper stack, and Frizzle handles the rest, from identification to grading to analytics generation.

Daily Formative Assessment for Individual Teachers

A middle school math teacher with five class periods can photograph each stack of daily quizzes or homework assignments and receive detailed analytics within minutes. The teacher sees which students are stuck on specific concepts, which misconceptions are emerging across the class, and which standards need reteaching. This allows the teacher to adjust tomorrow's lesson plan based on actual student performance data rather than intuition or delayed grading. The time saved from manual grading, estimated at 10-15 hours per week, can be redirected to individual student support or lesson planning.

Standards-Level Coaching and Professional Development

Instructional coaches can use Frizzle's aggregated data to run specific, evidence-based conversations with teachers. Instead of generic coaching sessions, coaches can review classroom-level misconception patterns, identify which teachers are struggling with particular standards, and provide targeted professional development. For example, if data shows that 40% of students across a grade level are making distributive property errors, the coach can organize a workshop specifically on that misconception. This transforms coaching from anecdotal to data-driven.

District-Wide Curriculum and Equity Analysis

District administrators can use Frizzle to monitor math performance across all schools in real time. The platform's equity dashboards reveal performance gaps between demographic groups, schools, or classrooms as they emerge, rather than waiting for spring assessments. Administrators can evaluate which curricula are producing the best mastery rates for specific standards, identify schools that need additional resources, and track the impact of instructional interventions. This creates a live nervous system for math instruction at scale.

College-Level Math Course Support

In higher education settings, such as the pilot programs at Vanderbilt University and Arizona State University, Frizzle is used to grade handwritten problem sets in college math courses. The system handles the increased complexity of college-level mathematics while maintaining its 97% accuracy rate. Instructors receive detailed analytics on which concepts the entire class is struggling with, enabling them to adjust lecture pacing and office hour focus. This use case demonstrates Frizzle's scalability from elementary through post-secondary mathematics.

Frizzle is free for individual teachers to start using. There is no credit card required to begin. The platform offers a free tier for individual teachers. For schools and districts, pricing is available upon request and is designed to scale with the number of classrooms and students. Frizzle is currently live in over 30 schools and districts, including pilots at Vanderbilt University and Arizona State University. Specific pricing tiers for institutional plans are not publicly listed and require contacting the sales team.

How does Frizzle handle different handwriting styles and messy work?

Frizzle's computer vision model was trained on 1.4 million pages of authentic K-12 student work, which includes print, cursive, scribbled, and sideways handwriting. The system is designed to recognize the messy, partial, and varied ways real students write. It does not require students to print neatly or follow a specific format. If the system's confidence in grading falls below a certain threshold, it flags that paper for human review, ensuring that ambiguous work is not automatically misgraded.

Is Frizzle compliant with student privacy regulations like FERPA and COPPA?

Yes, Frizzle is fully compliant with FERPA (Family Educational Rights and Privacy Act) and COPPA (Children's Online Privacy Protection Act). The platform undergoes annual SOC 2 Type II audits to verify its security controls. Student work data is encrypted using AES-256 at rest and TLS in transit. Critically, student work is never used to train or improve the underlying AI models, ensuring that individual student data remains private and is not repurposed.

Does Frizzle require students to use tablets or digital devices?

No, Frizzle is explicitly designed to work with paper. Students continue to write with pencils on standard paper worksheets, quizzes, or tests. There are no student logins, no tablets, and no digital migration required. Teachers simply photograph the completed paper stacks using a phone, document camera, or scanner. Frizzle then reads the handwriting and links each page to the correct student automatically. This preserves the existing classroom workflow.

What standards and curricula does Frizzle support?

Frizzle supports the Common Core State Standards (CCSS), Texas Essential Knowledge and Skills (TEKS), and over 30 additional state frameworks. The system is curriculum-agnostic, meaning it can read and analyze work from any math curriculum, including Eureka Math, Illustrative Mathematics, Saxon Math, and others. The platform maps 147 named misconceptions to specific standards, enabling granular analysis of which standards each student has mastered and where prerequisite gaps exist from earlier grade levels.