A School Built Around the Experiment Ledger
Neurantis records what it teaches, what works, and what does not. This page is that record.
Back to HomeHow Neurantis Came to Exist
Neurantis was founded in Kuala Lumpur in 2021 by a small group of engineers who had spent years working in ML research and production systems across Southeast Asia. The founding observation was straightforward: engineers with strong implementation skills often struggled to extract a working system from a research paper, and engineers who could read papers often could not evaluate whether a method would survive contact with real data.
The gap between the literature and working code is not a mystery. Papers are written to communicate a finding, not to reproduce one. Hyperparameters get rounded. Ablations get cut. Failure modes get omitted. An engineer reading a paper for the first time does not know what the authors left out because the authors did not list it. Neurantis was built to address this directly, by treating the process of reading and reproducing a paper as a skill that can be taught and assessed.
The name comes from the idea of a neural ledger—an accumulated record of experiments, each with a hypothesis, a method, a result, and a note on where the result disagreed with the hypothesis. This structure runs through every programme Neurantis offers. A learner who finishes one of the courses leaves with a documented record of their own attempts, not just a statement that they completed a module.
Neurantis operates from Level 15, Menara Prestige in Kuala Lumpur and currently runs three programmes: a reading and reproduction course for engineers new to the ML literature, a reinforcement learning course for those already comfortable with deep learning, and a research engineering fellowship for practitioners who want to conduct and publish their own work.
The school's cohort statistics—completion rates, median hours, resubmission counts—are shared openly with prospective learners. The figures are not selected for their appeal. They are shared because a learner making a decision about a seventeen-week or twenty-four-week commitment deserves an accurate picture of what completing it actually takes.
The People Behind the Ledger
Siti Rahmah
Co-Founder · Research Lead
Siti spent eight years at two Malaysian universities designing ML curriculum before concluding that course completion certificates were measuring the wrong thing. She oversees all programme design and runs the RL office hours.
Arif Kamaruddin
Co-Founder · Engineering Director
Arif ran production ML infrastructure at a Kuala Lumpur fintech for six years. He wrote the original code review rubric used across all Neurantis programmes and supervises the fellowship track.
Li Wei
Senior Instructor · Papers Course
Li Wei completed a doctorate on neural architecture search and has reproduced, or attempted to reproduce, over forty published papers. She runs the weekly reading sessions and designs the reproduction exercises for the papers course.
How We Hold Ourselves to Account
Preregistered Experiments
Fellowship projects require a preregistration document before any large compute run begins. The hypothesis and evaluation criteria are fixed before results are visible.
Data Privacy Compliance
Learner data is processed under Malaysian PDPA requirements. Contact information collected during enrolment is used only for programme administration and is not shared with third parties.
Individual Code Review
Every assignment submission receives written feedback from an instructor, not an automated rubric alone. Review comments are stored and remain accessible after the cohort ends.
Ethics Review Process
Every fellowship project undergoes a structured ethics and data-provenance review before compute is allocated. The review is part of the programme, not optional or deferred.
Published Cohort Outcomes
Completion rates and median time-on-task figures are shared publicly each cohort cycle. The numbers are reported as they are, not filtered for appeal.
External Critique
Fellowship work is reviewed by two invited external reviewers before public release. Reviewers are drawn from industry and academic roles unaffiliated with Neurantis.
What Neurantis Knows and How It Works
The ML education market in Malaysia tends toward short courses that cover the surface of a framework rather than the reasoning underneath it. Neurantis sits in a different position. Its programmes are longer, ask more of a learner per week, and are oriented toward engineers who already have a foundation and want to go further in a specific direction.
The papers course is relevant to engineers who have found that following a tutorial on an ML technique is straightforward, but reading the paper the technique came from is not. The course teaches the reading and reproduction skills directly, using actual published papers as the material.
The RL course covers the theoretical and practical foundations of sequential decision making at a level suitable for engineers who intend to work with these methods in production or research environments. It does not promise a portfolio project that can be showcased on a resume. It covers the material in depth, including the evaluation problems that make RL results difficult to compare across papers.
The fellowship is the most demanding of the three and is meant for engineers who have already decided they want to contribute to the research record rather than only consume it. The structure—preregistration, supervised experimentation, external critique, public release—mirrors what a research group would expect, not what a course provider typically offers.
Neurantis does not place learners in jobs, does not promise specific outcomes, and does not issue qualifications recognised by any external body. It offers structured, supervised programmes of study and practice, with honest records of what learners did and how they performed.
See Which Programme Fits Your Background
Send an enquiry and we will match your experience to the right entry point.
Open an Enquiry