Participant feedback
▸ testimonials

What participants said
after completing the course

Reviews are from engineers in Malaysia who attended individual courses or employer cohort programmes. Names and roles are used with permission; identifying employer details are omitted on request.

120+

Engineers trained

4.6

Average course rating

14

Employer cohorts completed

3+

Years running

// reviews

Participant reviews

NF

Nurul Fadhilah

Data Analyst, Petaling Jaya

Python for ML Practice

The readiness quiz was the most useful part before I signed up — it showed exactly where my gaps were, and the first few weeks addressed those directly. The notebooks were harder to write than I expected. Not because the material was unclear, but because the grading rubric pushed you to document your thinking, not just produce output. I found the environment management module particularly practical; it fixed some bad habits I had picked up on my own.

June 2025

RS

Rajeswaran Subramaniam

Software Engineer, KL Sentral

Deep Learning Systems

I had tried to self-study backpropagation twice before and got stuck at the same point. Working through it in a structured session where I could ask questions and see it connected to the code was different. The lab report format took adjustment — I am used to writing comments in code, not structured reports — but by the third one it felt natural. The compute sandbox worked reliably; I never had to debug infrastructure when I should have been debugging training code. The reproduction task was genuinely difficult and worthwhile.

May 2025

AH

Asmah Hamzah

Research Associate, Shah Alam

Python for ML Practice

I appreciated that they told me upfront the course would be slow for my level based on the quiz. I enrolled anyway because I wanted to fill specific gaps in pandas, and that ended up being the right call. The plotting module alone was worth it — I had been using visualisations for decoration rather than debugging, and the distinction made a real difference to how I work now. One note: the weekly exercise time was closer to seven hours for me than five.

June 2025

LW

Lim Wei Jian

Engineering Lead, Kuala Lumpur

Employer Cohort

The scoping workshop at the start was time well spent. We had been vague internally about what we wanted to build, and the workshop forced us to be specific. The code reviews were useful in a different way than I expected — less about catching errors, more about surfacing design decisions that had been made implicitly and should have been made deliberately. The midpoint architecture document was blunter than I expected about what we should change, which I respected. The two clinics after close were short but addressed the questions that came up once we started using what we built.

May 2025

YB

Yusri Bakri

ML Engineer, Cyberjaya

Deep Learning Systems

The section on distributed training was more thorough than anything I had found on my own. Profiling was covered in a way that connected to actual training behaviour, not just as a list of tools. What I found most useful was that every claim about why something works referenced a paper I could read. When I disagreed with a point, I could go to the source. That is not common in courses I have taken before. The reproduction task was hard, and I did not fully match the paper's results, but the process of figuring out why was the most educational part.

June 2025

PT

Priya Thangaraj

Junior Developer, Bangsar South

Python for ML Practice

I came in knowing some Python but nothing about NumPy beyond what I had picked up from Stack Overflow. The course filled in what I was missing in a way that made sense of things I had been doing by imitation. The data-cleaning project at the end used a dataset that had real problems in it — not cleaned-up toy data — so the work felt like actual preparation. I would have liked a bit more time on vectorisation, but the notebook on it was thorough enough to revisit later.

July 2025

// case_studies

Detailed accounts

case_study_01.txt — Fintech team, KL // Employer Cohort

Challenge

A team of eight engineers at a Kuala Lumpur fintech company had started building a credit scoring model using an external ML library. After six months, they could run the model but had no documentation of why certain architecture choices had been made, and were unable to explain the model's behaviour to the compliance team. Two engineers had left, taking institutional knowledge with them.

What we did

The scoping workshop established that the core problem was design documentation rather than model performance. The twenty-week cohort ran the deep learning syllabus alongside weekly code review of the existing model. The midpoint architecture review identified three design decisions that needed to be revisited and documented. The final deliverable included an architecture document that the compliance team could reference.

Outcome

The team completed the cohort with a documented model architecture and a set of evaluation notes covering known limitations. The compliance review proceeded with the documentation as source material. Three engineers enrolled in the individual Deep Learning Systems course in the following intake.

"The midpoint review was harder to read than expected. But it was accurate, and it gave us something to work from."

— Engineering Lead, fintech company

case_study_02.txt — Individual progression // Python → Deep Learning

Challenge

A backend developer in Shah Alam had read several deep learning textbooks but struggled to connect the theory to working code. The concepts made sense in isolation but not when combined in a training loop. Attempts to implement backpropagation from scratch had stalled at numerical precision issues.

What we did

After taking the readiness quiz, she enrolled in Python for ML Practice to consolidate the numerical computing foundations. Eight weeks later, she moved directly into Deep Learning Systems. The backpropagation module, covered in the fourth week of that course, addressed the specific gap she had identified — connecting the mathematical form to the code directly, with a lab task that required implementing the gradient calculations without an autograd library.

Outcome

Completed both courses over six months, with a total of twenty-two weeks of study time. The reproduction task in the deep learning course used a 2022 attention-based architecture paper. She documented the deviations clearly, which the instructor flagged as one of the stronger reproduction submissions in that cohort.

"The entry course felt slow in weeks one and two. By week five I understood why it had been slow. I needed those weeks."

— Backend developer, Shah Alam

// contact

Reach the school directly

Address

68 Jalan Kemuja
59000 Bangsar, KL

Office Hours

Mon–Fri: 9 am–6 pm
Sat: 10 am–2 pm

// credentials

Professional standing

MDEC EdTech Directory 2024

Listed in the Malaysia Digital Economy Corporation's EdTech directory for structured technical curriculum delivery in Malaysia.

PDPA 2010 Compliance

Data handling practices for enrolment and assessment records are in accordance with Malaysia's Personal Data Protection Act 2010.

KL ML Practitioners Network

Active participation in the Kuala Lumpur ML Practitioners reading group, which predates the school's formal founding.

See the syllabus before deciding

The course pages describe what each programme covers, how it is assessed, and what the weekly workload looks like. Send an enquiry if you have questions.