LEANWARE TEAM
1 x Senior Full Stack Developer
AI Grader for Crowns
COMPANY
Custom Software Development
SERVICE
United States
COUNTRY
Fixed Scope
engagement MODEL
CLIENT OVERVIEW
Evan Menke from the University of Colorado (United States), was developing a platform to enable AI-powered evaluation of dental crown preparations. What began as a tool for his personal use quickly grew into a broader idea:
“What if this could support students across multiple dental schools?”
He partnered with Leanware to build a backend service capable of analyzing images of crown preparations with high accuracy. The goal was to help students self-evaluate their work and compare results alongside instructor feedback — improving learning outcomes and saving time for both students and faculty.
.png)
Backend: FastAPI with LangGraph for LLM workflow orchestration, PostgreSQL
Deployment: GitHub Actions → Render.com via Docker images
Version Control: GitHub
Project Management: Fixed-scope tracking with regular check-ins
Tech Stack Involved
.png)
Backend Implementation
Leanware delivered a backend system that integrates with multiple AI providers — including OpenRouter, OpenAI, Google, and Anthropic. The architecture allows the client to switch models seamlessly and compare accuracy across providers.
Highly Configurable Evaluation System
The backend enables full customization of:
Base prompts
Rubrics
Evaluation criteria
Score weights
This level of flexibility allows the client to adjust the grading system, fine-tune outcomes, and expand the solution as requirements evolve.
The backend service also may return a set percentage-based coordinate highlights for each evaluated image, which can help to visualize the feedback provided.
Integration with Base44
The backend was successfully integrated with Base44, the service hosting the AI Grader’s front-end UI. This ensured secure, reliable communication between the UI and the evaluation engine.
Initial Model Evaluation & Reporting
To determine which models performed best, Leanware conducted initial accuracy tests across different providers. A comparative report was delivered outlining performance, limitations, and recommended configurations.
Documentation & Code Quality
The solution includes clear configuration documentation and clean, maintainable code. This ensures future scalability, easier onboarding, and support for additional features or potential model training.
SERVICES PROVIDED
.png)
UX & UI DESIGN
.webp)
Leanware successfully delivered a backend service capable of evaluating images of crown preparations with high reliability. Evan has been able to run evaluations, store feedback for comparison, and iterate on his rubrics as new data is collected.
The collaboration may continue as the client gathers more evaluation samples, enabling enhancements to scoring criteria, and possibly training a custom model specialized in assessing dental rubrics, increasing accuracy and consistency for students across dental programs.
From Blueprint to Delivery
RESULTS
.png)
We love to take on new challenges, tell us yours.
We'll get back to you in 1 day business tops
