About Synthia
A Studio Built Around the Work of Learning
We set out to create a space where people could study AI seriously — without pressure, without hype, and with enough support to actually finish.
Back to HomeOur Story
How Synthia Came to Be
Synthia started in 2021 when a small group of machine learning engineers and educators in Chiang Mai found themselves asking the same question: why were most online AI courses either too shallow to be useful, or so dense that learners dropped out before the interesting parts?
The answer they kept coming back to was pacing and structure. A learner who spends three weeks genuinely understanding how to prepare a dataset will do far better work later than someone who skimmed five courses in a month. So we designed Synthia around that idea: fewer shortcuts, more substance, and someone to look over your shoulder at the right moments.
We chose to work in English because the field itself runs in English, and because we wanted to serve learners across Southeast Asia who were already working in that language. The studio operates from Nimmanhaemin Road in Chiang Mai — a neighbourhood long associated with creative and technical work — and all instruction, feedback, and support are handled from there.
Mission
To give anyone with curiosity and patience the tools to work with AI meaningfully — starting from data and building through to models they can explain and stand behind.
Vision
A community of developers and researchers in Southeast Asia who think carefully about what their models do and why — and who can communicate that clearly to others.
Values
Honesty in evaluation, patience with learners, clear documentation, and a preference for understanding over speed. We would rather a learner take twice as long and actually grasp the material.
The People
Who Runs the Studio
Nattapong Kriengkrai
Curriculum Director
Former researcher at a Bangkok data science lab, Nattapong designed the track progression and writes the core lesson materials for all three courses.
Siriporn Phromrak
Lead Mentor
Siriporn coordinates the mentor team and handles project reviews for the Generative Models track. She has a background in NLP and evaluation methodology.
Arthit Tangsomboon
Platform & Operations
Arthit keeps the learning platform running, manages enrolments, and is usually the first person learners hear from when they get in touch.
How We Work
Standards We Hold Ourselves To
Structured Curriculum Design
Every track is reviewed before launch and updated annually. Lesson objectives are explicit, and the sequence is tested with real learners before it goes live.
Written, Substantive Feedback
Project reviews are written by humans, not auto-generated. Each review addresses code, approach, and at least two specific suggestions. Turnaround is five working days.
Data Privacy & Security
Learner data is collected only for account and enrolment purposes, stored with encryption, and never shared with advertisers or third-party marketers.
Responsible AI Curriculum
Evaluation methods, documentation requirements, and honest reporting practices are part of every track — not optional extras. Learners leave with habits that hold up under scrutiny.
Regular Content Updates
The field moves fast. Core materials are reviewed twice a year and updated where the underlying methods or tooling has changed significantly.
Accessible & Inclusive Design
Platform and materials are designed for learners with different backgrounds and levels of prior experience. No prior degree is required to study at Synthia.
Our Approach
AI Education That Stands on Solid Ground
At Synthia, the word we return to most often is synthesis — the process of taking separate ideas and building something coherent from them. That describes what learners do in each track, and it describes how the school itself was put together: by drawing on pedagogy, engineering practice, and direct experience of what makes people give up on a course halfway through.
The data track exists because most learners who struggle with model work are actually struggling with what came before it: incomplete data, unclear features, or assumptions about the dataset that were never checked. Working through that material carefully, with exercises that produce something real, changes how people approach the rest of their work.
The model-building track was designed with experiment rigour in mind. There is a version of this subject that trains people to copy code and run it until something looks right. That version produces fragile work. The Synthia version asks learners to articulate why they tried what they tried, what the result means, and what they would change. Mentors are trained to ask those questions in their reviews.
The generative models track sits at the intersection of some of the most active research in the field and some of the most significant questions about how that research should be applied. The capstone project asks each learner to pick a problem they care about, work through it with the methods from the course, document what worked and what didn't, and present the result — not to impress, but to communicate clearly. That's the standard we hold ourselves to as well.
See Which Track Fits
Have a look at the full course offerings, or get in touch and we'll help you figure out where to start.