We teach the engineering side of working with AI
Neuralaya runs short courses and a part-time programme for developers who need to understand what they're building, not just that it runs.
Back to HomeHow Neuralaya came about
Neuralaya started in George Town in 2022, when a small group of developers noticed a gap between what ML tutorials covered and what teams here actually needed to do. The tutorials showed you how to call an API or run a notebook. They didn't show you how to evaluate whether a change in your prompt made anything better, or how to keep a serving system honest six months after launch.
The name comes from a compound of "neural" — as in the layer-based architectures that underpin the field — and "alaya", a Sanskrit-derived word present in several languages across the region, meaning place or repository. The intent was something like: a place where the technical side of this work can be understood properly.
The courses we run now are the ones we wish had existed when we were starting out. They're written for developers with professional experience in software, not for people who want a general overview. Prerequisites are stated plainly. Workload is given as hours per week, not "self-paced" or "intensive". What each course leaves out is listed alongside what it covers.
We are based at 72 Lebuh Chulia in George Town — in a building that's been used for various kinds of technical work for decades. Sessions are held live online. We think being in Penang matters for the cohort culture and for access; the programme is designed so that learners across Malaysia, Singapore and the wider region can attend without relocating.
What we're here to do
- Describe our courses honestly — syllabus, prerequisites, workload and scope — so you can decide whether they're worth your time before joining.
- Teach the engineering concerns that shape how ML systems behave in practice, not just the modelling concepts that dominate most reading material.
- Give feedback that's written by people who build systems, in enough detail to be useful rather than encouraging.
Things we won't do
- Describe a course as "beginner-friendly" if it isn't.
- Leave out what a course doesn't cover, so the syllabus looks more complete than it is.
- Make claims about what learners will achieve or earn after completing a course.
- Use language that sounds impressive without meaning anything specific.
The people who run the courses
Ahmad Nadzri
Lead Instructor, ML Engineering
Spent seven years building data and serving infrastructure at companies in Kuala Lumpur and Singapore before moving to teaching full-time. Designed the ML Engineering Programme syllabus and reviews all capstone projects.
Priya Letchumanan
Course Author, Language Model APIs
Worked on developer tooling and API integration at a software consultancy in Penang. Wrote the API course materials from scratch after finding that existing resources skipped evaluation almost entirely.
Wei Jun Tan
Instructor, Fine-Tuning Intensive
Runs the weekend intensive on fine-tuning. Previously worked in NLP research and has spent the last few years adapting open-weight models for domain-specific tasks in practical production settings.
How we hold our own work to account
These aren't aspirational values — they're the specific practices we use when writing course materials, running sessions, and reviewing learner work.
Syllabus review cycle
Course materials are reviewed before each new cohort starts. If something in the field has changed enough to affect what learners should know, the syllabus is updated — not kept the same for the sake of consistency.
Prerequisite checks
The ML Engineering Programme requires a short technical conversation before enrolment. This is how we make sure the fit is right for both sides — not to filter people out, but to avoid situations where a learner joins a course that isn't what they needed.
Written review standards
Project reviews go line by line. Reviewers use a written guide to ensure consistency across cohorts. Vague feedback ("good effort", "needs improvement") is not acceptable under our internal standards.
Data and privacy
Learner data is used to operate the courses and nothing else. We do not share personal details with third parties, do not add people to mailing lists without explicit opt-in, and keep data only as long as needed for record-keeping purposes.
Cohort feedback process
Each cohort completes a structured review at the end of their course. We read every response and use it to decide what to change. The feedback form asks specific questions rather than a general "how was it" rating.
No inflated outcomes
Our materials describe what the course covers and what learners will produce. We make no statements about employment, income or career progression. A technical course is a course, not a career contract.
Technical AI education in Malaysia
The field of machine learning has grown quickly, and so has the volume of material written about it. Most of that material targets two groups: people who want a general sense of what AI is, and researchers who need deep theory. Developers who write production software and want to understand how to use these tools reliably fall between the two.
Neuralaya addresses the middle ground. The API course covers the mechanics of calling hosted models reliably — prompting, structured output, tool calling, evaluation — which are the things a working developer needs to understand before building anything serious on top of a language model. The fine-tuning intensive covers dataset construction and parameter-efficient adaptation for developers who need to specialise a small model for a narrow task without large-scale infrastructure.
The ML Engineering Programme sits at a longer timescale. It covers the operational concerns that determine whether an ML system keeps working after the initial build: pipelines, versioning, serving, monitoring, and the habits that catch drift and degradation before users do. Project work makes up about 70 percent of the time, and reviews are written by engineers who have dealt with these problems in real systems.
Penang has a long history in the technology sector in Malaysia, and George Town in particular has a density of technical talent that isn't always visible in the typical conversation about the industry here. Neuralaya is part of that local ecosystem rather than separate from it.
Questions about how we work?
Send us a message. We'll give you a straight answer rather than a sales pitch.
Send a Message