Yam Studio
PySpark End-to-End ebook
A 113-page guide in Bahasa Indonesia that takes learners from first concepts to production patterns.
- DataFrames, transformations and joins
- Performance tuning and partitioning
- Common pitfalls, with fixes
Practical data engineering, taught from real project work.
Yam Studio turns hands-on experience with PySpark, dbt and GCP into structured lessons, ebooks and courses for learners in Indonesian and English.
Bootstrapped, founded in 2026, based in Jakarta, Indonesia.
Lesson: avoid collect() on large dataPySpark
# Slow: pulls every row onto the driver rows = df.collect() total = sum(r["amount"] for r in rows) # Fast: let the cluster aggregate total = df.agg(F.sum("amount")).first()[0]
Every lesson starts from a real mistake seen in client work, then explains why it happens and what to do instead.
Learning data engineering is harder than it needs to be
The problem
Material is scattered across docs, blog posts and videos. Most of it explains syntax, not the decisions engineers make on real projects. Structured material in Indonesian is especially rare, so beginners stall between tutorials and the job.
Our approach
We start from what actually happens in production work and teach it in order: concepts, patterns, common mistakes and how to fix them. An AI pipeline helps us produce and update material quickly, and every piece is reviewed by a working data engineer before it is published.
What we make
One paid product in market, three channels publishing every day, and two products in development.
Free daily channels
Produced end to end by our agent pipeline and published on a schedule.
- PySpark and data engineering tips in Indonesian, 3 posts a day
- AI agents and automation in English, 3 posts a day
- YouTube Shorts on AI updates and vibe coding, 2 videos a day
Courses and a learner assistant
Structured courses on dbt and GCP built from our full note library, plus a Q&A assistant that answers learners' questions using our own material.
How we build
A self-hosted, multi-agent system runs our whole publishing operation. More than ten specialised agents run on scheduled jobs, and every run reports to a Telegram workspace where we review, adjust or cancel before anything goes out.
1
Research
Agents collect sources: our own data engineering notes, official release notes and changelogs, and community discussions.
2
Write and check
A writing agent drafts posts, scripts and chapters. A second agent scores each draft and sends weak ones back for a rewrite.
3
Produce
Images, voice-over, subtitles and short videos are generated and rendered automatically.
4
Publish and learn
Content is scheduled to Threads and YouTube. An analytics agent reads post insights and picks the next day's topics.
Built on Claude. Claude, served through Amazon Bedrock, handles the writing, reviewing and analysis stages. The agents run in Docker on our own hardware, and our public site, legal pages, media delivery and email run on Cloudflare (Pages, Tunnel and Email Routing).
Progress so far
-
Aug 2026
Studio founded
Started building the agent pipeline and the first learning materials.
-
Sep 2026
First paid product
The 113-page PySpark ebook, sold on Lynk.id with a free sample.
-
Oct 2026
Analytics-driven publishing on Threads
Two accounts at 3 posts a day each. An analytics agent reads post insights and chooses the next day's topics.
-
Oct 2026
Automated YouTube Shorts channel
Research, scripting, review, voice-over, rendering and scheduled upload, 2 videos a day, with no manual editing.
-
Next
Structured courses
dbt and GCP courses generated from the full note library and reviewed chapter by chapter.
-
Next
Learner Q&A assistant
An assistant that answers questions grounded in our own lessons.
About
Yam Studio was founded by Yosie Abdul Muzanil, a data engineer who works hands-on with Spark on client projects. The studio started as a way to organise years of working notes, and became a way to share them with people learning the field.
The goal is simple: make practical data engineering easier to learn, and use AI for the repetitive work so the teaching stays focused and accurate.
We also offer a Spark performance audit for teams whose Spark jobs are slow or expensive. It draws on the same hands-on experience behind our learning material.
- Founder
- Yosie Abdul Muzanil
- yosieam@yamstudio.fyi
- Store
- lynk.id/yosieam
- Location
- Jakarta, Indonesia