Qognito courses.
Decide, build, explore: choose your course.
Do you need to assess an AI investment, develop agents for your business or clients, or prepare your students for these practices?
Qognito courses cover three complementary areas: decision-making and governance, agent engineering, and the mechanisms of reasoning and learning.
Three complementary courses
Choose the course that matches your needs and experience.
SAGA-IA
AI Agent Architecture & Governance Strategist
What should you delegate to AI, at what cost and within which limits? Learn to assess a project, compare options and justify your decisions. For leaders and managers who commit resources, from independent businesses to large organisations.
No programming background required
BOOTAG-IA
Agentic Harness Engineering
Learn to build AI agents, connect them to tools and knowledge, then monitor and evaluate how they work.
For developers, technical consultants, engineers and students who want to understand and control the mechanisms behind their applications.
REASON-IA
Reasoning, Emergence & Reinforcement Learning
Explore how models produce responses and how different methods can improve their results. Experiment with reasoning, evaluation and reinforcement learning to assess their performance and limitations.
For AI professionals, R&D teams, specialist consultants and advanced students.
Which course fits your project?
Are you looking to develop your skills, train your team or incorporate these topics into a curriculum?
Tell me about your context and the skills you want to develop. We can discuss the relevant course, available materials and suitable arrangements.
Discuss my project →SAGA-IA: AI Agent Architecture & Governance Strategist
Invest in AI with decisions you can justify.
What should you delegate? At what cost? Within which limits? SAGA-IA helps you assess a project, question a proposal and decide whether to proceed, narrow the scope or stop.
For independent professionals, owners of one- or two-person businesses, business managers, and executive, finance or IT leaders. No programming skills are required.
Discuss my SAGA-IA project →Five steps to make decisions and retain control
- Section 1
Choose what to delegate
- Compare delegation with an option that does not use AI and set conditions for proceeding.
- Understand why corrections recur before choosing a remedy.
How to put your decision to the test
A reasoned decision and a diagnosis of rework: facts, assumptions, possible causes and missing evidence.
- Section 2
Assess full costs and viability
- Compare full costs at equivalent quality and service coverage.
- Calculate the conditions for viability, including an adverse scenario.
How to put your decision to the test
A cost comparison and a conditional decision. Freed capacity is distinguished from savings that can actually be realised.
- Section 3
Reuse what is worth keeping
- Choose what should become documentation, a procedure, a control or a program.
- Plan tests, ownership, version management and withdrawal of these reusable assets.
How to put your decision to the test
A reuse decision and lifecycle plan, tested against cases where the solution should work and cases where it should not apply.
- Section 4
Set limits on what AI can do
- Identify potential harm involving data, access and permitted actions.
- Specify protections and assess tests of refusal, approval and recovery.
How to put your decision to the test
Verifiable requirements and a decision accounting for remaining risks. An instruction alone does not restrict an actual permission.
- Section 5
Roll out gradually and keep alternatives open
- Set responsibilities, quality checks and conditions for expansion or suspension.
- Assess dependencies and plan how to resume the service with another solution.
How to put your decision to the test
A rollout plan and recovery exercise with costs and acceptance criteria. An exportable file alone does not guarantee continuity.
Practise decisions through concrete cases
Two contexts run through the course: business intelligence in an organisation and a one- or two-person business. These cases are simulations unless an observation is explicitly documented. Each step requires a justified choice; gathering more evidence or stopping can be the right decision.
Learning is assessed through the reasoning and evidence presented. Taking the course alone does not guarantee gains, savings or the compliance of a product or service.
What do you, your team or your programme need?
Describe your context and needs. We can discuss available materials, relevant sections and arrangements for a teaching session.
Discuss my SAGA-IA project →Course in preparation: the first two sections have been written. The final syllabus and presentation materials are still being developed.
BOOTAG-IA: Agentic Harness Engineering
28 hours · Four 7-hour days or staggered sessions
For developers, technical consultants, engineers and students who want to understand and control the mechanisms behind their applications.
Prerequisites: Advanced object-oriented programming practice in R or Python, proficiency with REST APIs and JSON, and foundational prompting skills.
1. What you will learn to do
Build the execution environment
Connect a model to tools, organise its context and define the steps through which the agent works.
Diagnose and repair
Use execution traces to investigate network failures, context loss and orchestration errors.
Assemble and evaluate
Combine the components into an executable harness, monitor resource use and test its behaviour from end to end.
2. Build, test and understand each component
The harness is the code that organises model calls, tools, memory and controls. You build it progressively without an orchestration framework, to understand architectural choices and intervene when a component fails.
Workshops use R and R6 with direct REST API calls. They connect context management, knowledge access, tools and coordination between agents.
Learning through practice
Debugging exercises challenge you with deliberately faulty code. The capstone involves assembling a complete harness, producing execution traces and running it through an evaluation suite.
BOOTAG-IA for you, your team or your curriculum
Explore the learning sequence, exercises and assessment methods. Tell me about your context so we can check how the course fits your experience, team or curriculum.
Send me the programme for BOOTAG-IA →REASON-IA: Reasoning, Emergence & Reinforcement Learning
40 hours · Five 8-hour days
For AI professionals, R&D teams, specialist consultants and advanced students.
Prerequisites: Basic programming skills in R or Python and an elementary understanding of deep learning: neural networks, backpropagation and PyTorch.
1. What you will learn to do
Understand response generation
Experiment with text generation, the model’s computational memory and different strategies for selecting responses.
Measure before comparing
Build a verifiable evaluation process and compare methods for reasoning, voting and correcting responses.
Train and examine trade-offs
Experiment with reinforcement learning and distillation into a compact model, examining accuracy, stability and compute requirements.
2. Explore how models work through experimentation
You implement the mechanisms under study to observe their effects: generation, evaluation, response improvement and training. The aim is to understand when a method helps and how much computation it requires.
The course combines R for analysis and experiment control with Python and PyTorch for computation, through reticulate. Topics include reinforcement learning with verifiable rewards (RLVR, using GRPO) and distillation.
A project that puts your choices to the test
Exercises lead to a project involving the training and distillation of a compact model on accounting or financial logic problems. You justify evaluation, reward and infrastructure choices using observed results.
REASON-IA for you, your team or your curriculum
Explore the learning sequence, exercises and assessment methods. Tell me about your context so we can check how the course fits your experience, team or curriculum.
Send me the programme for REASON-IA →