India has over 70 million people with dyslexia. Most of them are students. Almost none of the digital exam resources they use were designed with them in mind. Lexara Air started as a question: what if an AI could do the adaptation work automatically?
The Problem Is Structural, Not Personal
When students with dyslexia struggle with a question paper, the common assumption is that the student needs more practice. The actual problem is that the document was never designed for them. Dense walls of text, inconsistent fonts, no spacing hierarchy, English-only — these are structural barriers, not cognitive ones.
Lexara Air does not change the student. It changes the document. Every exam PDF that enters the pipeline comes out the other side re-formatted, re-staged, and re-paced for a learner whose brain works differently.
The 6-Stage Cognitive Scaffolding Pipeline
The core insight behind Lexara Air is that comprehension happens in stages, not all at once. A complex question cannot be answered until it is first understood — and understanding requires breaking it down.
The pipeline works as follows: OCR extracts the raw question text from the PDF. GPT-4 Vision then segments the question into a 6-stage scaffold: keyword identification, concept anchoring, step-by-step breakdown, worked example, self-check prompt, and answer framing. Each stage is a checkpoint, not a paragraph.
// Pipeline principle
“Never show a student the full question until they have already worked through its components. Comprehension is sequential. The UI should enforce that sequence.”
Why 9+ Indian Languages
English is the medium of instruction in most Indian engineering exams. It is not the first language of most Indian engineering students. This creates a dual barrier: the cognitive load of the subject matter plus the cognitive load of processing a second language.
Lexara Air addresses this by integrating a translation layer that can output each scaffold stage in Hindi, Tamil, Telugu, Marathi, Bengali, Gujarati, Kannada, Malayalam, Punjabi, and Odia. The student reads the concept in their language, then answers in English. The barrier is halved.
Adaptive Typography as Accessibility Infrastructure
OpenDyslexic is a typeface designed to anchor characters visually so they are harder to flip or invert. Lexara Air applies it by default. But typography is only part of the picture. Line spacing set to 1.8, paragraph width capped at 65 characters, and a high-contrast dark-background mode collectively reduce the visual noise that dyslexic readers find exhausting.
None of these are cosmetic choices. Each one has a measurable effect on reading speed and error rate for dyslexic learners — effects that are documented in the education accessibility literature.
What This Project Changed About How I Build
Before Lexara Air, I thought of accessibility as a checklist. ARIA labels, contrast ratios, keyboard navigation. Important, but downstream of the design.
Lexara Air taught me that real accessibility is an upstream architectural decision. The pipeline structure, the language model prompting strategy, the scaffold sequencing — these are all accessibility decisions made before a single line of UI code is written.
The best time to design for accessibility is before the product exists.