What people say after taking a Neuralaya course
Feedback from developers across Malaysia and Singapore who've gone through the API course, the fine-tuning intensive, and the ML Engineering Programme.
Back to HomeLearner feedback
Reza Asyraf
Backend Developer, Kuala Lumpur
I joined the API course because my team was adding an AI feature and I kept running into evaluation problems — we couldn't tell if our prompts were actually better or just different. The section on eval was the part I needed. The evaluation harness is still something I use on my own projects.
LM API Course · June 2025
Tan Li Ying
Software Engineer, Penang
The weekend fine-tuning intensive was more technically demanding than I expected — particularly the dataset construction day. I came in thinking that was the straightforward part. Leaving with a trained adapter was satisfying. The written review of the eval report was useful, though I'd have liked more time on Day 2.
Fine-Tuning Intensive · May 2025
Muhammad Firdaus
Full-Stack Developer, Johor Bahru
The ML Engineering Programme was the right decision for where I was — I could write Python and had shipped a few data pipelines, but had no idea about drift monitoring or how to think about a model registry. Twenty weeks is a real commitment but the project reviews were worth it on their own. The feedback on my second project ran to about four pages.
ML Engineering Programme · Cohort 4, 2025
Siti Khadijah
Data Engineer, Selangor
I appreciated that they told me the programme wasn't quite right for me the first time I enquired — I didn't have enough Python experience. I spent two months working on some side projects, came back, and that conversation was still being taken seriously. The technical call before enrolment is straightforward, not a test.
ML Engineering Programme · Cohort 5, 2025
Kevin Chai
Platform Engineer, Singapore
The API course is compact. Four sessions, but each one moved. I had been building on language model APIs for about six months before taking it and there were still things I hadn't thought about properly — mainly around graceful degradation and the cost accounting. Recommended for anyone who's been doing this ad hoc.
LM API Course · July 2025
Nurul Ain Razak
ML Engineer, Kuala Lumpur
Went into the fine-tuning weekend with a small classification dataset from work that I hadn't been able to get much traction on. The day-one session on dataset quality explained why. Came out with a working adapter and a clearer picture of what "working" means when you're doing narrow task adaptation. Worth the trip from KL.
Fine-Tuning Intensive · June 2025
A closer look at two learner journeys
From ad hoc prompting to a measured pipeline
Challenge
A backend developer at a SaaS company in KL had spent three months building an AI-powered document summarisation feature. The team kept arguing about whether prompt changes were improvements. There was no way to settle it objectively — every iteration was evaluated by feel.
Course
He joined the LM API course specifically for the evaluation section. Over four weeks, he built the harness alongside the course exercises and adapted it to work against their existing test documents. By week three he had a score function that could compare prompt versions automatically.
Outcome
The team now runs prompt experiments against the harness before shipping. The arguments about "feel" stopped because there's a number. The harness is the course deliverable he adapted — he has owned it since the last session.
"The evaluation section was the only reason I joined and it was worth it. Everything else was useful too, but that was the gap I came to fill."
— Reza A., LM API Course
Building the operational half of an ML system
Challenge
A developer at a logistics company in Penang had trained a demand forecasting model that worked well in notebooks. Getting it into production, keeping it there, and knowing when it was drifting were a different problem. She had the modelling; she didn't have the engineering.
Programme
She joined the ML Engineering Programme and spent 20 weeks on the parts she'd been missing: versioned pipelines, a model registry, a serving layer, and monitoring that would catch drift in her feature distributions. Her three reviewed projects built toward the capstone — a refactored version of the production system.
Outcome
The capstone review identified two design decisions in her pipeline that the instructor would have done differently, with the reasoning written out. Those were the most useful five pages of the whole programme — not because they were critical, but because they were specific and arguable.
"The project reviews were what made it worth the time. Not the sessions, not the recordings — the written reviews. That's where the actual feedback was."
— Programme learner, Cohort 3
Reach us
Phone
+60 4-261 9473Address
72 Lebuh Chulia
10200 George Town
Pulau Pinang
Office hours
Mon–Fri: 10:00–18:00
Sat: 10:00–14:00 MYT
By the numbers
3
Years running courses in George Town
230+
Developers trained across the region
4.6
Average satisfaction (out of 5, cohort surveys)
100%
Programme learners receive written capstone review
[note, numbers, context]
The satisfaction figure comes from end-of-cohort surveys that ask specific questions rather than a general rating. Response rates vary by cohort; we don't average across cohorts with significantly different response rates. We publish what we can stand behind.
Looking for more information before deciding?
Send us a message with your situation and we'll describe which course makes sense, or say honestly if none of them fit right now.
Send a Message