Corvidla
AI course programmes

Course Programmes

Three programmes, one coherent arc from foundations to research

Each course covers a distinct layer of AI development work. They can be taken sequentially or chosen according to where you currently stand in your learning.

Back to Home

Sequential or standalone

The three programmes build on each other naturally — mathematics, then architectures, then research — but each can also be entered directly by learners with the relevant background. Prerequisites are listed plainly for each course.

Fixed timelines

Courses run on fixed schedules, not at your own pace. A cohort starts, works through a defined programme, and ends. This structure exists because it is the condition under which most learners actually finish.

Built around working knowledge

The measure of whether the teaching has worked is not whether you can recall a definition but whether you can use the idea in a new context — read a paper that relies on it, implement something that requires it, or identify when it does not apply.

Programme 01

Mathematics for Machine Learning

8 weeks 5–6 hrs/week RM 590

An eight-week cohort covering the mathematics that modelling work actually uses, taught with code alongside every derivation. Topics include vectors and matrices as transformations, eigendecomposition and why it matters for dimensionality, gradients and the chain rule, probability distributions, expectation and variance, maximum likelihood, and information-theoretic measures. Each idea is implemented in Python immediately after it is derived. Intended for learners who can code but stalled on the mathematics when reading papers or documentation.

What's included

  • Eight live sessions with full replays
  • Weekly problem sets with worked solutions
  • Fortnightly derivation clinic
  • Cohort forum for questions between sessions
  • One mentor call included
  • Completion record on finishing required work

Prerequisite

Comfortable writing Python. No specific mathematics background required beyond secondary-level algebra.

Enquire About This Course
Mathematics for Machine Learning

Typical weekly flow

01

Live session: concept introduced, derivation worked through, Python implementation shown

02

Problem set released: three to five questions requiring derivation and implementation

03

Submission reviewed by instructor, comments returned within two working days

04

Derivation clinic (fortnightly): common mistakes addressed, alternative approaches discussed

Transformer Architectures in Depth

What you will build

01

Tokenisation and embedding layer from scratch in Python

02

Self-attention and multi-head attention with positional encoding

03

Full training loop with normalisation choices, then fine-tuning and adaptation techniques

04

Paper reproduction project: end-to-end implementation of a published result at small scale

Programme 02

Transformer Architectures in Depth

24 weeks 12–14 hrs/week RM 2,940

A twenty-four-week cohort dedicated to sequence modelling and attention-based architectures, built up from scratch. Learners implement tokenisation, embeddings, self-attention, multi-head attention, positional encoding, normalisation choices and a full training loop, then move to fine-tuning, adaptation techniques, quantisation, inference optimisation and evaluation design. Considerable time is given to reading current papers critically and reproducing a published result at small scale. Aimed at engineers and researchers who want working knowledge rather than API familiarity.

What's included

  • Twenty-four live sessions with replays
  • Compute credits for experiments and the reproduction project
  • Weekly code review with specific feedback
  • Paper reproduction project assessed by instructor
  • Four mentor calls spaced through the programme
  • Final architecture defence and detailed completion record

Prerequisite

Solid Python skills. Comfortable with the material from the mathematics programme (eigendecomposition, gradients, probability). The mathematics course is the recommended entry path.

Enquire About This Course

Programme 03

Research Practice Programme

32 weeks 13–16 hrs/week RM 4,660

A thirty-two-week programme for learners moving toward research or applied research roles. It covers experimental design, ablation discipline, statistical treatment of results, reproducibility practice, writing for peer review, poster and talk preparation, and the ethical review questions that accompany work involving people or scraped data. Each participant carries one research question through the whole programme, from literature survey to a written paper and a presented talk assessed by working researchers. Suited to those with solid modelling foundations and the patience for slow, careful work.

What's included

  • Thirty-two weeks of teaching, seminars and review sessions
  • Compute credits for the research project experiments
  • A supervisor assigned for the full programme duration
  • Fortnightly writing clinics
  • Full paper draft with two review rounds
  • Presented talk assessed by working researchers and detailed completion record

Prerequisite

Solid modelling foundations and familiarity with the implementation side of at least one architecture type. Experience reading papers is helpful but not required — literature survey methodology is covered in the programme.

Enquire About This Course
Research Practice Programme

The research arc

01

Weeks 1–8: Literature survey, research question scoping, experimental design

02

Weeks 9–20: Experiments, ablations, statistical analysis, reproducibility checks

03

Weeks 21–28: Writing, two review rounds with supervisor and external reviewer

04

Weeks 29–32: Talk preparation and assessed presentation to working researchers

Decision Guide

Choosing the right programme

Maths
RM 590
Architectures
RM 2,940
Research
RM 4,660
Duration 8 weeks 24 weeks 32 weeks
Weekly hours 5–6 12–14 13–16
Live sessions 8 24 32
Mentor calls 1 4 Supervisor throughout
Compute credits
Paper reproduction
Research paper + talk

Best for

Mathematics course

Developers and data practitioners who can write Python but find ML papers hard to follow because the mathematics stops them.

Best for

Architectures course

Engineers who want to build and adapt transformer models rather than just call APIs, and who want to read current papers critically.

Best for

Research programme

Those with solid modelling backgrounds moving toward research or applied research roles, who want to produce and present original work.

Across All Programmes

Standards shared by every cohort

Data privacy

Submission materials and personal data are handled according to our Privacy Policy and are not used outside the programme context.

Marked by instructors

Every submission across all three programmes is reviewed and marked by an instructor. Automated grading is not used for any learning work.

Updated between cohorts

Materials are reviewed and updated after each cohort based on participant feedback and changes in the field. Version notes are shared with enrolled learners.

Honest prerequisites

Prerequisites are written to be accurate. Applicants who do not yet meet them are encouraged to address that first rather than entering a course under-prepared.

Not sure which programme fits?

Write to us briefly — your background, what you are trying to learn and why — and we will give you a direct answer about which course makes sense and when the next cohort opens.

Get in Touch