## core_benefits_summary
Six things we do that many AI courses do not
Exercises use untidy real data
Not curated teaching sets. Every data exercise begins with something that actually needs cleaning.
Written feedback, not rubric scores
A practising engineer reviews every submission line by line and explains the reasoning behind their comments.
Clear statements of what is not covered
Each programme explicitly lists its out-of-scope topics. You know the edges before you start.
Small cohorts
The seminar caps at 30. The development track is smaller. Office hours remain useful rather than performative.
Evening pacing, part-time scope
Sessions are in the evenings. The development track is designed for 12–15 hours a week alongside a full-time role.
You own the work you produce
Project repositories belong to participants when the track ends. No proprietary platform lock-in.
## benefit_detail_blocks
Expertise — teaching by practitioners
The people who teach at Gradient Hall work on AI systems professionally. Course content is not assembled from literature reviews; it reflects decisions that have been made and revised in production environments. Feedback on submitted work carries the same context.
- Instructors with backgrounds in data engineering, ML serving, and evaluation
- Capstone reviews conducted by a practising engineer who did not design the project
- Reading lists drawn from primary papers and technical documentation
Technology — working in the tools professionals use
The development track covers the transformer architecture end to end before moving to retrieval systems and tool-using agents. Compute credit is included so participants can run experiments rather than simulate them. Work happens in notebooks and version-controlled repositories, which is where the work happens professionally.
- Compute credit allowance included in the development track fee
- All exercises require actual code that runs, not pseudocode
- Notebooks used throughout — consistent with professional AI workflows
Support — access to instructors, not just recordings
Recordings are provided for every session, but the development track also includes fortnightly one-to-one office hours. These are not optional extras — they are part of the programme structure. The cohort channel is kept small enough to be genuinely useful for questions between sessions.
Pricing — clear, no hidden tiers
Fees are stated per programme, in Malaysian Ringgit, with a full list of what is included. There are no premium add-ons to unlock later. The data course is RM 980, the seminar RM 480, and the development track RM 4,650. What you see on the page is what the programme costs.
Outcomes — what completing the track produces
The development track ends with a self-defined capstone that must run end to end and survive a review from someone who did not build it. Four graded projects are reviewed before that point. The repository, the code, and the project history are yours when the programme concludes. We make no claims about employment or earnings.
## comparison_with_alternatives
How this differs from a typical AI course
| Feature | Typical AI course | Gradient Hall |
|---|---|---|
| Exercise datasets | Pre-cleaned teaching sets | Genuinely messy real-world files |
| Project feedback | Automated checks or rubric scores | Line-by-line written review by a practising engineer |
| Scope statement | What is covered, broadly | What is and is not covered, explicitly |
| Office hours | Community forum or none | Fortnightly one-to-one with an instructor |
| Compute access | Often at participant's expense | Credit allowance included in track fee |
| Prerequisites honesty | Vague or absent | Stated plainly; track includes a prior conversation |
| Claims about outcomes | Employment and salary claims common | No employment or earnings claims made |
## unique_aspects
Three things we do not see elsewhere
The evaluation seminar is a standalone evening
Most courses treat evaluation as one chapter. We built an entire seminar around it because it is where teams most often make expensive mistakes. It runs in three hours and is designed for people who read code but may not write it daily.
Dataset documentation template included in the data course
One of the outputs of the five-week data course is a dataset documentation template that participants can use on their own projects. Good documentation is the part of data work that makes a March result reproducible in September.
Capstone must survive a review from an independent engineer
The final project on the development track is reviewed by someone who was not involved in advising it. That constraint — building something a stranger can understand and run — is one of the more useful constraints to work within.
## school_milestones
Where the school stands
3
Structured programmes covering data, evaluation and full-stack AI
30
Seminar participant cap, keeping Q&A time usable
4
Graded projects reviewed by a practising engineer on the development track
2h
Response time target for enquiries, within working days
## enquiry_cta
Want to talk through whether a programme suits your situation?
Send a message and we will reply within two working days. No pressure to enrol immediately.