← ALL WORK
AI & education

bildim

AI assistance. Teacher judgment.

An education workspace connecting document understanding, assessment and teacher-controlled AI workflows.

NuxtVue.jsTypeScriptPythonFastAPIPostgreSQL
MY ROLE

Product & Full-Stack Engineering

BUILT FOR

Teachers managing learning materials and assessment

01 / THE CONTEXT

AI assistance. Teacher judgment.

Turning documents and student work into useful teaching material involves repetitive preparation and review. Automation is only useful when teachers retain control over what becomes part of the learning experience.

What I built

I connected a Nuxt workspace with Python services for document and AI processing. Review and approval are part of the product workflow, so machine-generated output can be evaluated before it is used.

02 / PRODUCT CAPABILITIES
01

Document understanding

OCR and vision processing extract information from educational documents and feed structured downstream workflows.

02

Teacher-controlled output

Review and approval steps keep educators involved in AI-assisted material creation and assessment workflows.

03

Background processing

Celery and Valkey separate expensive document processing from the interactive teaching workspace; ONNX supports local model execution paths.

Under the surface.

Follow a real product workflow through its application boundaries. Select any step to explore the implementation.

ENGINEERING IN MOTION

Capture. Interpret. Let the teacher confirm.

A focused view of handwritten score capture and rubric-assisted evaluation, where machine output remains reviewable.

STEP 01 / OWNERSHIP CHECK

Exam context

The API verifies the teacher owns the exam and associated records before starting a capture or confirmation workflow.

WHY IT MATTERS

The important boundary is between a candidate interpretation and a confirmed educational record. Different material workflows have their own approval semantics.

Conceptual flow based on the implementation. Animated connections illustrate data movement, not live traffic.

04 / ENGINEERING JUDGMENT

Built around the constraints.

Treat AI output as material to review. Separate compute-heavy processing from interactive requests, and make teacher approval a visible part of the workflow.

Discuss the engineering ↗

Keep exploring.

Next project ↗