Campus · District · Aquin Co-pilot

When the professor says 'let me show you'.

Aquin is the real-time teaching co-pilot embedded into every AquinTutor lecture hall. It listens to the professor, infers intent, and generates diagrams, animations, code, and 3D models in the side panel within 200ms. The professor previews, approves, projects. Students get a perfect replay — the artefact stamped to the exact second it was conjured.

200ms

Median latency

Modalities

Indic languages

On-device

When possible

What Aquin does.

Six capabilities that turn a passive whiteboard into a generative surface. None of them require the professor to leave the flow of the lecture — Aquin runs alongside, the way a skilled TA would.

Live diagrams

Vector diagrams generated mid-sentence. Block diagrams, circuits, flow charts, organic chemistry skeletons, free-body diagrams — all rendered as SVG that can be paused, labelled, and re-played.

On-demand 3D

Atoms, molecules, lattice structures, proteins, planetary orbits, mechanical assemblies. Orbit with mouse, slice with key, isolate sub-structures. Pure WebGL fallback to canvas where unsupported.

Math typesetting

Spoken equations parsed and typeset in 80ms. Greek letters, summations, integrals, bra-ket notation, tensors with super/subscripts, matrix layouts. Hand-rolled glyph layout — no MathJax dependency.

Code that runs

Aquin writes Python, JS, C, SQL, Verilog, QASM. Code blocks have a Run pill that executes in a sandboxed worker. Output (numeric, plot, animation) appears inline. Compatible with the campus IDE.

Multimodal Q&A

Students point a phone at a textbook page, a circuit, a botanical leaf, a Sanskrit shloka. Aquin reads, interprets, and answers in the student's chosen Indic language with audio + text + supporting diagram.

Personal replay

Every Aquin artefact (diagram, code run, animation) is timestamped against the lecture audio. Students scrub back to "the moment Aquin drew the Bloch sphere" and step the visual frame-by-frame.

Live demo.

Click any prompt on the left. The stage on the right renders the artefact in your browser — no server roundtrip, no library. Same renderer the professors use in lecture.

Bloch sphere

Renderer: SVG vector · Latency: 84ms

theta (polar) 0.00 rad

phi (azimuth) 0.00 rad

|psi> = 1.000 |0> + 0.000 |1>
⟨X⟩ = 0.000 ⟨Y⟩ = 0.000 ⟨Z⟩ = 1.000

The pipeline.

From lecturer's lips to projected artefact in roughly 200 milliseconds. Each stage runs on the lecturer's local rig where possible, with a stateless fallback if the on-device renderer can't handle the artefact.

Professor speaks

budget: 0ms

Lecture mic captures audio at 48kHz. Audio stays on the lecturer's device — never uploaded as bytes.

Speech-to-text

budget: 60ms

On-device streaming ASR. Indic-tuned acoustic model with subword decoder. Partial hypotheses emitted every 20ms.

Intent classification

budget: 40ms

Renderer

budget: 100ms

Vector / 3D / animation / code-eval / typeset pipeline. Output appears in the side panel for the professor to preview, approve, drop into the lecture.

Faculty controls.

Aquin never bypasses the professor. Every artefact lands in a preview pane; the faculty member decides what reaches the class. Edits flow back as training signal so Aquin improves with the faculty member's style.

Preview

Every Aquin output lands in a side panel first. The professor sees the artefact before the class does. Nothing auto-projects without faculty consent.

Approve

A single keypress (space / pedal) pushes the artefact to the main screen and into the student replay log.

Edit

Aquin output is editable in-place. SVGs are real elements, code is real text, math is re-typesettable. Faculty corrections become Aquin training signal.

Save to replay

Approved artefacts are stamped to the lecture timeline so any student can rewind weeks later and watch that exact diagram appear in context.

Back to campus districts Try the labs

Privacy by default.

No audio, no video, no raw classroom media leaves the room. Aquin earns its keep by being a guest in the lecture hall — not a surveillance device.

On-device transcription

Audio never leaves the lecturer's device. ASR runs locally on the campus rig (Apple Silicon / consumer GPU sufficient).

No media bytes uploaded

Only the transcript fragment and the renderer-output artefact (text, SVG, code) leave the device. Raw audio + raw video stay local.

Student opt-out

Students may flag a clip as private; the corresponding transcript fragment is purged from replay within 30 minutes.

Faculty owns artefacts

Diagrams Aquin draws under a faculty member's direction belong to that faculty member. License terms shown in the publish dialog.