Skip to main content
Axiya

§1 — The free tool

Accessibility
infrastructure
for STEM.

Drop a STEM PDF — handwritten lecture notes, scanned textbook chapters, equations, figures. Get back accessible HTML with screen-reader-navigable math.

Free for individual use. Built toward something larger (see §4).

Sign in (free) to start.

Open signup with any verified Google account. Your transcriptions live on your dashboard, ready to download as accessible HTML.

Sign in to startSee a sample first

§2 — What the tool produces

A page of handwritten lecture notes, transcribed in plain HTML.

BeforeSource PDF · page 1 · scanned handwriting
Photograph of a page of handwritten lecture notes for an MIT 6.334 power-electronics class. The page shows hand-drawn equations for Kirchhoff's Current Law, integrals defining time-averaged currents, a small node diagram of three branch currents meeting at a point, and a numbered boxed list of five rules for a power converter in periodic steady state.
AfterAxiya output · structured HTML · screen-reader navigable

Page 1 · verbatim

6.334 Lecture Notes: Intro to DC/DC

Small hand-drawn node diagram showing three labelled branches converging at a single point, with arrows indicating direction.

Brief alt

Small hand-drawn node diagram showing three labelled branches converging at a single point, with arrows indicating direction.

Detailed figure description

A small hand-drawn node diagram showing three straight branch lines converging at a single central junction. One branch comes from the upper-left, one from the lower-right, and one from the lower-left; each branch has a short arrow near the junction pointing toward the junction. Each arrow is labeled with a Greek lambda: the upper-left branch is labeled \(\lambda_n\), the right branch is labeled \(\lambda_1\), and the lower-right branch is labeled \(\lambda_2\). There are small dotted marks drawn near one branch to indicate continuation or additional branches beyond the three shown. The sketch is compact and positioned near the top-right of the page.

First Averaged circuit rules

KCL

\[ \sum_d \lambda_d = 0 \]
\[ \frac{1}{T}\int \sum_d \lambda_d \, dt = 0 \]
\[ \sum_d \frac{1}{T}\int \lambda_d \, dt = 0 \]
\[ \sum_d \langle \lambda_d \rangle = 0 \]

KCL applied to time-average branch quantities (cons. charge).

The same is true for KVL.

\[ \sum_k \langle V_k \rangle = 0 \]
So for a power converter in Periodic Steady State:
  1. Average KCL
    \[ \sum_d \langle \lambda_d \rangle = 0 \]
  2. Average KVL
    \[ \sum_k \langle V_k \rangle = 0 \]
  3. Capacitor in P.S.S.
    \[ \langle \lambda_C \rangle = 0 \]
  4. Inductor in P.S.S.
    \[ \langle V_L \rangle = 0 \]
  5. If system lossless (cons. of energy)
    \[ \langle P_{in} \rangle = \langle P_{out} \rangle \]

§3 — Pipeline

Four steps, start to finish.

  1. Upload

    Any STEM PDF — handwritten lecture notes and scanned textbooks both go through the same path. The pipeline reads each page as an image.

  2. Transcription

    Equations rendered via MathJax with assistive MathML on. Figures get a brief alt and an expandable detailed description — short alt for quick orientation, long description for everything a sighted reader sees.

  3. Verification

    Every output runs through axe-core, Pa11y, and Lighthouse before delivery. They verify what is machine-detectable: structure, contrast, heading order, alt presence, keyboard navigability.

  4. Edit

    Refine equations, descriptions, and reading order in plain English. No LaTeX knowledge required — this is where Axiya stops being a transcription tool and starts being a pipeline.

§4 — Where Axiya is going

Maxiya.

Manual STEM accessibility remediationcosts universities $7–12 per page and takes days per document. The pool of workers who can do this credibly — STEM-literate, accessibility-trained, capable of handling math and figures — is in the low thousands globally. Demand is billions of pages a year by 2027. The arithmetic doesn’t work, and what’s actually happening is that faculty are pulling materials offline rather than risk personal liability under Title II.

Not a claim that AI can do it all. A claim that the labor can collapse by orders of magnitude.

No open-domain AI system has delivered reliable 100% accuracy on a non-trivial task in the last several decades, and accessibility won’t be the first. The honest picture: an agent handles the long tail of pages end-to-end, and a small team of STEM-literate human reviewers handles the genuine edge cases — the non-linear figures, the unusual notation, the documents that need real judgment.

The shape of the systemis direct integration with the LMS. A faculty upload into Canvas, Blackboard, Brightspace, or Moodle hits the pipeline automatically — transcription, figure descriptions, accessibility audit, edge-case escalation to the reviewer team, return to the LMS — in minutes. Faculty do not touch a remediation queue, do not fill in a request form, do not learn an interface. The accessibility office sees status across every document at a glance. The reviewers touch only what the agent flagged as uncertain.

Roughly a hundred reviewers, instead of a few thousand. Billions of pages a year, instead of a tiny fraction of demand. Every output carrying an audit trail an institution can defend.

That is what Maxiya is being built to do. It is in early development and not yet generally available. If you’re an accessibility office or institutional buyer interested in being among the first to use it, get in touch.

Email shonu@axiya.ioFor institutional pilots

§5 — Get in touch

Two paths. Pick the one that fits.

For faculty + individual users
The free tool sits at the top of this page. Drop a PDF and the conversion starts immediately.
↑ Back to the tool
For accessibility offices, administrators, institutional buyers
If you’re evaluating Axiya for use across a department, school, or system — or you want to be among the first to pilot Maxiya — email me directly.
shonu@axiya.io ↗