We teach AI development by working through the code line by line

Synapsia started from a specific observation: developers working with AI tools in Kuala Lumpur understood what the tools produced but not how they worked. That gap was not about effort or intelligence — it was about access to material that went past the API call and into the mechanism itself.

The school was put together in 2022 by a small group of engineers who had spent years reading the literature and writing implementations from scratch. The founding idea was that understanding a transformer well enough to build one, even a small one, changes how you read every subsequent paper, every library and every failure mode. The school should teach that kind of understanding, not shortcuts around it.

The first cohort ran the Foundations course in early 2023, working through tokenisation, attention and the training loop across six weeks of evenings. The written feedback experiment — reviewing each submission in writing rather than in a quick verbal note — turned out to matter more than expected. Students referred back to those written responses weeks later, when the same issue came up in a different context.

The Deployment short course followed in late 2023, addressing a gap the Foundations cohort identified themselves: the distance between a working notebook and a service that other people depend on is larger than it looks, and the practical details — batching, warm starts, logging — are not covered in most ML courses.

The Full Practitioner Programme launched in 2024 for developers who wanted to work through the complete arc: models from scratch through adaptation, retrieval, agents, evaluation, serving and a capstone they define and hand to a reviewer who did not build it. Roughly seventy per cent of the programme is project work, reviewed line by line by practising engineers.

Synapsia is based at 19 Jalan Cangkat Raja Chulan in Kuala Lumpur. Enrolment starts with a conversation, not a purchase.