Kuala Lumpur · Founded 2021
We built Nyxlearn to show the full fabric before asking anyone to enrol.
Three AI development programmes, each with its prerequisites, workload, and assessment method stated from the first paragraph.
Back to HomeOur Story
How Nyxlearn came to exist
Nyxlearn was set up in Kuala Lumpur in 2021 by a small group of software engineers who had spent years answering the same question from junior colleagues: "Which course should I do to get into machine learning?" The honest answer was usually "none of these, exactly" — most available courses either glossed over the mathematics or dropped learners into deep theory without a working programming context.
The name comes from the idea of weaving: a loom creates fabric from threads that cross at defined points. The dependency graph of AI knowledge works in a similar way. Gradient descent only makes sense once you have a mental model of a loss function. A transformer architecture is harder to debug if you do not have a clear picture of tokenisation. Nyxlearn's courses are structured around that dependency order, and the syllabus makes those dependencies explicit before asking anyone to pay.
We started with the Foundations of Machine Learning course, running the first cohort of eighteen developers through twelve weeks of part-time live sessions. Feedback from that cohort shaped the workload estimates, the assignment structure, and the decision to cap each intake at twenty-four. The Applied Language Model Engineering course followed in 2023, built for developers who had completed a foundations course and wanted to work specifically with large language models in a production context.
The Team Capability Programme was added after several Kuala Lumpur-based engineering teams contacted us asking whether the individual courses could be run internally. The programme is different in structure: it begins with a capability review of the team rather than a fixed syllabus, and the code review clinics run against the team's own repositories.
Nyxlearn is not a degree-granting institution. We issue a written record of completion when a learner finishes a course and meets the assessment criteria. We do not claim accreditation and do not make assertions about employment outcomes. What we do claim is that the syllabus is published in full, the prerequisites are named, and the workload estimate is based on what previous cohorts actually reported spending, not on what we hoped they would need.
The People
Tutors and programme leads
Each course is delivered by a tutor who has built and shipped the kind of system being taught, not just read about it.
Amir Karim
Programme Lead · ML Foundations
Previously a data engineer at a Kuala Lumpur fintech. Teaches the Foundations course and writes the graded assignments. His background is in gradient-based optimisation applied to risk modelling.
Siti Wan
Tutor · LLM Engineering
Spent three years building retrieval-augmented systems for a regional logistics firm. Leads the Applied Language Model Engineering course, with particular focus on evaluation and red-teaming.
Razif Noor
Lead · Team Capability Programme
Works with engineering teams on the Team Capability Programme. Background in MLOps and deployment architecture. Conducts the initial capability review and designs the tailored syllabus for each team cohort.
Our Standards
How we run our courses
Full syllabus before enrolment
Every module, its topic, and its position in the dependency order is published before any enrolment step. You decide what you are signing up for with the complete picture in hand.
Graded, not just completed
The Foundations course uses three graded assignments and a written reflection. The LLM course uses a deployed project plus a defence session. Completion without assessment is not the standard.
Tutor reads every submission
Cohort caps exist so this is possible. The twenty-four seat limit on individual courses is not a marketing device — it is the number at which each tutor can give substantive feedback on every piece of work.
Data handled with care
Enquiry data is used only to respond and, if you enrol, to administer your course place. We do not sell or share data with third parties for marketing purposes. Full details are in our Privacy Policy.
Sources named where it matters
When the curriculum covers safety and governance practice for language models, every claim references a named publication or organisation. We do not teach opinion as established fact.
Honest about what a record means
Completion records are written records of attendance and assessed work. They are not academic qualifications, carry no accreditation, and we do not make claims about how employers will treat them.
AI development courses designed around how the knowledge fits together
The Foundations of Machine Learning course covers the mathematical and statistical tools that practising engineers actually use, rather than a broad survey of every technique in the field. Linear algebra is taught in the context of how weight matrices behave during training, not as a standalone topic. Probability is introduced through the lens of how a model expresses uncertainty about its outputs. This framing means a developer who finishes the course can read a paper about a new loss function and understand why the authors chose it.
The Applied Language Model Engineering course picks up where foundations ends. It covers the architecture of transformer-based models in enough depth to reason about fine-tuning behaviour, then moves quickly into the practical work: how to budget for token costs across a production API, how to write evaluation scripts that catch regressions, and how to document a retrieval-augmented system so another engineer can maintain it. The module on safety and governance practice draws on published sources from organisations that research these questions professionally.
The Team Capability Programme is structured differently because engineering teams have different needs from individual learners. The starting point is a written capability review: what tools the team currently uses, which parts of the AI stack they want to own versus integrate, and what a realistic internal project would look like. The weekly sessions and code review clinics are then shaped around those answers. The curriculum handover at the end of the programme means the team has documented what they learned and how, in a form they can share with new joiners.
All three programmes are delivered online, with live sessions scheduled to fit working hours in the MYT timezone. Kuala Lumpur-based participants can also visit the Jalan Pinang office for any component they prefer to attend in person, subject to prior arrangement.
Questions about how we work?
Send a message and we will answer clearly. No obligation, no follow-up call unless you ask for one.
Get in Touch