Corvidla
Learning environment

Why Corvid Labs

What you get here that you do not find elsewhere

The gap between knowing how to call a library and understanding what it is doing is large. These courses are built to close it.

Back to Home

At a Glance

The core advantages of studying with us

First-principles teaching

Concepts are derived before they are applied. You finish knowing why something works, not just that it does.

Code in every session

Python implementations follow immediately after each derivation, keeping theory and practice aligned throughout.

Fixed cohort schedule

Clear start and end dates with weekly deadlines. Structure matters for retention and completion.

Specific written feedback

Every submission is marked by an instructor. Comments address your specific approach, not a model answer.

Small cohorts by design

Cohort sizes are kept small enough that quality feedback and instructor attention are sustainable.

Sample sessions freely available

Full recorded lessons are accessible before enrolment. Decide based on what you actually see, not what we say about it.

Expertise

Instructors who work at the level they teach

The instructors at Corvid Labs have backgrounds in applied research and engineering, not just curriculum delivery. The lead instructor for the mathematics and architectures programmes spent several years in compute research before moving into teaching. The research programme director holds a background in statistical learning theory. This matters because the questions learners bring — about edge cases, about why a particular result does not hold in a specific setting, about the assumptions behind a paper — are questions that need someone with working knowledge to answer.

  • Questions answered with reference to underlying mechanisms, not documentation
  • Feedback written by someone who has solved the same problems
  • Curriculum decisions made by practitioners, not by committee

Teaching philosophy

"A learner who can re-derive a result from scratch understands it. One who can only reproduce a code block from memory does not. We teach toward the former."

— Corvid Labs teaching guidelines

What's included technically

Python problem sets with worked solutions
Compute credits for cloud experiments (architectures and research)
Session replays available after each live session
Cohort forum for asynchronous discussion
Reference implementations reviewed by instructors

Technology

Modern tools, current methods

Courses are taught using the tools and frameworks that appear in current research and production work. The architectures programme covers quantisation and inference optimisation techniques that have become relevant only in the last few years. The research programme addresses the evaluation and reproducibility concerns that have emerged as the field has scaled. Materials are updated between cohorts when the relevant methods have shifted enough to matter.

Support

Responsive and substantive

Mentor calls are scheduled within the programme fee for all three courses. The research programme includes a supervisor assigned for the full thirty-two weeks. Responses to forum questions from instructors arrive within two working days. We do not use automated responses for learning questions — if a question needs a real answer, a person writes one.

  • Mentor calls included in course fee (one call for maths, four for architectures)
  • Dedicated supervisor for the research programme
  • Fortnightly derivation clinic and writing clinics by programme

Value

Transparent pricing, nothing hidden

Course fees in Malaysian Ringgit cover everything listed in the programme description — sessions, replays, problem sets with worked solutions, mentor calls, compute credits where applicable, and the completion record. There are no separate materials charges or platform subscription fees. The fee structure is set before enrolment opens and does not change mid-cohort.

Mathematics for Machine Learning RM 590
Transformer Architectures in Depth RM 2,940
Research Practice Programme RM 4,660

Outcomes

What you leave with

Ability to read papers

After the mathematics programme, the notation and derivations in standard ML papers are readable rather than opaque. The architectures course goes further, to the point where recent transformer papers can be reproduced at small scale.

Working implementations

The architectures cohort ends with participants having implemented a full transformer, trained it, and applied fine-tuning and quantisation techniques. These are not demo notebooks — they are implementations that can be extended.

Research output

The research programme produces a written paper draft, two review rounds and an assessed talk. Participants who complete it have submitted work for external review and have defended it in front of working researchers.

Comparison

How we compare with typical alternatives

A factual look at what is commonly offered elsewhere and what this programme provides instead.

Feature Typical online course Corvid Labs
Derivations taught from first principles
Code implementation alongside theory
Specific written feedback on submissions
Fixed cohort with real start and end dates
Mentor calls included in fee
Compute credits included
Full sample session before purchase decision
Paper reproduction project (architectures)

What Sets Us Apart

Features you will not find in similar programmes

Misconception panels in every topic

Each area of the curriculum includes a list of the mistakes learners most often make at that point, written with brief corrections. These are not generic warnings — they come from seeing the same errors in submissions across multiple cohorts.

Paper reproduction at small scale

The architectures programme includes a project where participants reproduce a published result, end to end, at a compute scale that fits within the included credits. Reading a paper critically and then implementing it is a different skill from following a tutorial, and this is where that skill develops.

One research question, start to finish

The research programme is built around a single question that each participant carries through the full thirty-two weeks, from literature survey to a written paper to a presented talk. The coherence of that arc is something that collections of separate assignments cannot replicate.

Full sessions as a pre-enrolment sample

We make complete recorded lessons available before any payment is made, not excerpts selected to highlight pace and clarity. If the teaching style is not right for you, you can discover that before committing to the programme.

Track Record

Milestones since we opened

6

Completed cohorts

140+

Learners graduated

11

Research papers produced

94%

Cohort completion rate

Malaysia Digital Economy Corporation — Partner School

Recognised as a partner institution for technical AI upskilling, 2024

IEEE Educational Affiliate — Northern Region

Affiliate status with IEEE Malaysia Section, education committee, 2025

Top-rated AI school — Penang Tech Community, 2025

Voted by members of the Penang Tech Community forum, July 2025

See a sample session, then decide

We make a complete recorded session available before any enrolment discussion. Get in touch and we will send you the link for the course you are considering.

Request a Sample Session