Automate publishing
Fiction, STEM textbooks, and bilingual courses—director-to-reviewer agent tiers produce book-length output with parallel writers and retry-aware completion.
AI automation · Education · Agentic systems
Automating everything—including education
AI agent systems that automate publishing, learning, and knowledge work—from novels and textbooks to full language courses and the workflows behind them.

Our theme
When each agent tier completes its work, the next domino falls—drafting, reviewing, exercising, and answering at scale without sacrificing coherence.
One domino triggers the next. We build hierarchical AI agents that turn manual creative and educational pipelines into repeatable, local-first automation—so humans steer vision while machines draft, review, and scale.
Fiction, STEM textbooks, and bilingual courses—director-to-reviewer agent tiers produce book-length output with parallel writers and retry-aware completion.
Curriculum outlines, lesson sections, typed drills, and model answer keys— language courses with vocabulary tables, exercises, and reviewer gates built in.
Custom orchestration, benchmarks, and on-prem inference stacks for teams that need agentic automation without cloud lock-in.
Watch
A quick introduction to our mission—automating publishing, education, and knowledge work with hierarchical AI agents.
Products and research that automate publishing and education—from agent-orchestrated books to curriculum pipelines—built as composable, local-first AI on GitHub Pages.
AI-powered parody generator with syllable matching, rhyme schemes, semantic coherence, and local LLM generation via Ollama—so songs become themed parodies that keep the original rhythm and style.
Test your knowledge of Bible verses in an interactive quiz game—start a round, tune settings, and climb the top scores leaderboard.
Guess the Prompt—an interactive challenge game built around a curated corpus of AI prompts. Tune settings, play rounds from the hive menu, and sharpen prompt literacy by reverse-engineering what instruction produced each output.
Turn any reading passage into quizzes automatically—on-device Llama 3.2 via Ollama keeps generation local so curriculum teams can draft assessment items without sending content to the public internet.
Automate end-to-end book production—fiction novels, STEM textbooks, and bilingual language courses—with hierarchical agents that plan, write, review, and generate exercises plus answer keys on local LLMs.
AI-native curriculum generation—outlines, lessons, drills (fill-in-the-blank, multiple choice, matching), and separate reviewer tiers—so instructional design scales like software deployment.
Indexes spoken media into pause, sentence, word, and syllable clips, then composes requested sequences with an inverted index that prefers unused takes and ranks reuse by usage—research product with an industrial-track paper; packaged binary shipping when the release is ready.
Industrial-track papers and benchmark galleries documenting how agentic automation performs on real book-length and education workloads.
Design automation for your domain—education platforms, publishing houses, or regulated ops—with orchestration, evaluation harnesses, and on-prem inference.
Industrial research on automating long-form publishing and education— papers, benchmarks, and public showcases that prove agent tiers can replace manual drafting at scale.
Industrial research track · 2026
Hierarchical multi-agent automation for fiction, textbooks, and language education— director-through-reviewer tiers, parallel scene writers, exercise generators, and retry-aware completion that treats publishing and teaching as one automatable pipeline.

Leadership
Sole Technical Founder
Building Domino Data Systems around a single conviction: automate everything worth repeating—including how we teach, learn, and publish. From hierarchical multi-agent book generation to evaluation harnesses for local LLM pools, the goal is composable AI that ships as research-grade public showcases.
shyamalschandra.github.io