whiz.coach Engineering

White Papers

Architectural notes from the team building whiz.coach. Multi-agent pipelines, Google Cloud composition, and the engineering techniques that keep AI-generated educational content honest at scale.

Published

v1.0 · 2026-04-12 Ashish Awasthi Cloud generation + local validation

Visually Engaging Study Material Generation and Validation using Multi‑Agent Orchestration

How twelve specialised AI agents on Google Cloud, paired with a fully-local MLX validation tier, turn raw syllabus sources into responsive, audio-narrated, SVG-rich study material at scale.

Covers: syllabus planning, topic content generation, image and interactive SVG authoring, narration with Cloud TTS and forced-alignment, HTML rendering, validation tiers, local correction loop, prompt-feedback synthesis.

Read the paper
v1.0 · 2025-12-27 Rishit Awasthi Cloud Tasks agent pipeline

Pedagogically Rigorous Question and Diagram Generation and Validation using Multi‑Agent Orchestration

How five specialised AI agents on Google Cloud generate examination-grade questions, illustrate them with educationally faithful diagrams, and validate both through generic AI visual analysis with multi-layer integrity checks.

Covers: Bloom's-taxonomy-balanced question generation, GPT-class SVG authoring with Gemini fallback, multimodal validation against the rendered PNG, iterative diagram correction with delayed counter increment, the angle-arc three-tier fallback, learner-feedback loop.

Read the paper