LearningWrite / Education / AI product
AI-assisted writing practice for the classroom.
LearningWrite was an affiliated partner company owned by Mike Biglan. We helped build its language-writing practice around conversational scenarios, AI-assisted feedback and classroom workflows.
Explore the story ↓
01 / The challenge
Useful practice needs useful feedback.
Writing practice needs a useful prompt and feedback a learner can act on. Educators also need to assign activities, understand progress and fit the tool into an existing classroom environment.
More about the challenge
LearningWrite’s product challenge brought these needs together. AI could support an interaction, but the surrounding experience had to make that capability useful to students and teachers.
- Project
- LearningWrite
- Our contribution
- Research, product design, AI integration & engineering
- People using it
- Students, teachers & administrators
- Engagement
- Education / AI product
02 / A product decision
Autoscoring for instant feedback
The auto-scoring system used a fine-tuned language model built with PyTorch and Hugging Face and served through TorchServe. It connects written responses with feedback based on language-assessment frameworks, giving learners something to review within the practice flow.
03 / Inside the work
Give AI a purpose in the learning experience.
Research with language educators and program administrators informed the feedback, conversational-practice and integration work. The design considered different roles and the steps around an assignment.
The platform connects student practice with educator tools, bringing writing support and assignment workflows into the same product.
04 / A product decision
LMS integration
The work includes integrations with classroom and rostering systems for sign-in, roster imports and assignment workflows. These connections make adoption part of the product design rather than an afterthought.
05 / A product decision
Pre-written prompts, for assignments made simple
Work with the University of Oregon’s Center for Applied Second Language Studies informed the writing activity library. Activities can be found by proficiency level, topic and “can-do” statements to support assignment selection.
06 / A product decision
Role management ecosystem
Ed tech tools need to work for all of a school’s users. The LearningWrite platform offers specialized portals for students, teachers, administrators, and content managers, designed to meet the unique and overlapping needs of each role.
07 / A product decision
Multilingual writing support
Writing prompts and conversational scenarios support language practice in multiple languages. The learner needs a clear situation and room to respond; the educator needs an activity that fits a teaching goal.
08 / Our contribution
Language expertise.
A complete learning product.
We worked with language educators and program administrators on the product, and with the University of Oregon’s Center for Applied Second Language Studies on writing activities. Those contributions shaped what students practice and what educators can assign.
Twenty Ideas connected the AI scoring and conversational features with role-specific interfaces, sign-in, rosters and assignment workflows. Model integration was one part of building a product that could fit a classroom.
LearningWrite was used in multiple school districts, including Albuquerque Public Schools.
Under the surface: how this project was built
Scoring and serving
Auto-scoring used a fine-tuned language model with PyTorch and Hugging Face, deployed through TorchServe.
Classroom connections
Canvas LMS and ClassLink integrations supported the surrounding sign-in and classroom workflows.
Application foundation
React and TypeScript supported role-specific interfaces; Kotlin, Spring Boot and PostgreSQL formed part of the backend.
09 / Product work
Give writing practice a place in the classroom.
Conversational scenarios and AI-assisted feedback meet the workflows of students and educators.
Work with Twenty Ideas
Bring us the work
you want to make possible.
Tell us who it is for, where the experience breaks down and what needs to change.
How we build and evolve products