Engineering the practice of building AI systems
Three courses covering Python data engineering, transformer architecture, and capstone system design — for practitioners who want to work through the material rather than consume it.
Courses
Each course is a standalone unit with a defined syllabus, weekly sessions, assignment review, and a written record of what was completed.
Python & Data Engineering for AI Work
Ten weeks. Vectorised numerical code, memory profiling, columnar formats, reproducible environments, dataset versioning, and pipelines that fail loudly. Written for analysts and backend developers moving into model work.
- Nine assignments with written code review
- Dataset audit on real Malaysian open data
- Environment reproducibility check on every submission
- 6–8 hours per week expected
Transformer Architectures & Language Model Training
Eighteen weeks. Attention, tokenisation, positional schemes, pretraining objectives, supervised fine-tuning, and evaluation — implemented from the paper, not from a wrapper. Suited to engineers comfortable with PyTorch and calculus notation.
- Shared cluster access throughout
- From-scratch transformer trained on a public corpus
- Weekly paper-reading sessions
- 12–15 hours per week expected
AI Systems Engineering Capstone Track
Twenty-six weeks. Retrieval, serving, evaluation, monitoring, cost control, and error analysis. For practitioners who already train models and need the surrounding system to hold up under load. 15–20 hours per week expected.
- Dedicated technical mentor throughout
- Fortnightly architecture reviews
- Written incident post-mortem required
- Closing technical presentation
How the courses are structured
Each design decision reflects the kind of learning that holds up when you are six months into a real project.
Material from the source
Transformer architecture is taught by working through the paper, not through a library abstraction. You build what you study.
Code review on every submission
Each assignment receives written feedback on correctness, readability, and engineering choices — not just a pass/fail mark.
Cluster access included
GPU compute and inference budget are part of the course fee. Hardware constraints are documented honestly per assignment.
Known failure modes documented
Every course unit lists where learners commonly get stuck. You can read what goes wrong before it happens to you.
Responsible deployment covered
Dataset licensing, model risk, provenance, and consent are addressed as engineering considerations, not policy appendices.
Written completion record
Each learner receives a written assessment record documenting what was completed and how. All code written remains the learner's.
Have questions about prerequisites or course fit?
Send a message and someone from the team will read it and respond with a plain answer — no sales process, no pressure.
Common questions
Do I need a machine learning background to enrol in the first course?
What Python experience does the first course expect?
Is the Transformer Architectures course done entirely online?
How is the Capstone Track different from the Transformer course?
What is the total fee and how is payment structured?
Do I keep the code I write during the courses?
How is my personal data handled when I submit an enquiry?
Find Us
Suite 8-3, Menara Suezcap, KL Gateway, 59200 Kuala Lumpur
Enquire
Use the form below or contact us directly. We read every message and reply with a plain, useful answer.
Contact details
KL Gateway, 59200 Kuala Lumpur
Monday – Friday: 9:00 AM – 6:00 PM
Saturday: 10:00 AM – 2:00 PM
Closed on public holidays