Agentic
Harness
Engineering
Track°
How modern agent harnesses really work, and how to make them reliable on your own codebase and systems: skills, MCP tools, subagents, evals and guardrails.
Hands-on AI courses for teams that build and run AI: agentic harness engineering, rapid acceleration sprints, R&D with AI, the full software-to-hardware ecosystem, private RAG, governance and security.
Agentic
Harness
Anatomy
Models, tools,
loops
Models, tools, loops, permissions: how Claude Code, Codex and similar harnesses actually work.
Skills &
Memory
Reusable
know-how
Package team know-how as reusable skills, memory and project rules.
MCP &
Tool Design
Agents meet
your systems
Connect agents to your systems through MCP servers and well-shaped tools.
Subagents &
Orchestration
Planner, worker,
verifier
Planner / worker / verifier patterns, parallel agents, workflows.
Evals &
Verification
Proof before
shipping
Tests, evals and review gates so agent output never ships unchecked.
Guardrails &
Security
Safe by
default
Prompt injection, secrets, sandboxing and human approval points.
Teams move faster when they understand the tools they use. The next course turns that understanding into a working sprint.
AI Rapid
Acceleration
Team
Sprint°
Five days from "we should use AI" to a live agent on your data, a measured gain and a 30-day plan the team owns.
Day 1
Process Audit
Find the
wedges
Map recurring work and pick the three highest-leverage wedges.
Prompt to
Workflow
Repeatable,
documented
From ad-hoc chats to repeatable, documented AI workflows.
Build the
First Agent
Live, on
your data
Ship one working agent on your own data, live, in the room.
Measure
the Gain
An honest
ROI sheet
Time saved, error rate, cost per task: an honest ROI sheet.
30-Day
Adoption Plan
Make it
stick
Owners, rituals and guardrails that keep the change alive.
Speed matters most where knowledge is created. Research teams get their own track.
R&D
with AI
Research
Track°
Use AI to scan the field, widen the hypothesis space, simulate before building and keep every result reproducible.
Literature &
Patent Agents
Know what
is new
Agents that scan papers, patents and datasets and report what is new.
Hypothesis &
Experiment Design
Wider, then
sharper
Use models to widen the option space, then narrow it with evidence.
Simulation &
Digital Twins
Test before
you build
Test ideas in simulation before spending on hardware or field trials.
Data Pipelines &
Synthetic Data
Data for your
own models
Collect, clean, label and synthesise data for your own models.
Reproducible
Research
Rerun
anything
Notebooks, evals and audit trails that a reviewer can rerun.
Prototype to
Product
Paper, patent
or product
From a working result to a paper, a patent or a product.
To choose well you need the whole map. One course covers the stack from silicon to regulation.
The AI
Ecosystem
Software ↔
Hardware°
The whole AI stack in one course: from silicon and data centres to models, agents, robotics, cost and regulation. Overview first, then in depth.
Silicon
Where AI
physically runs
GPUs, TPUs, NPUs, memory bandwidth and why HBM decides speed.
Infrastructure
Cloud, on-prem,
edge
Cloud, on-premise, edge; inference servers, quantisation, latency.
Models
Frontier vs
open weights
Frontier vs open weights, fine-tuning, distillation, evaluation.
Data &
Retrieval
RAG, vectors,
graphs
RAG, vector stores, knowledge graphs and data governance.
Agents &
Applications
The new
software stack
Harnesses, MCP, workflows and the new software stack.
Robotics &
Embodied AI
The physical
frontier
VLA models, humanoids, ROS 2 and the physical frontier.
Economics &
Regulation
Costs, energy,
EU AI Act
Compute costs, energy, EU AI Act, market map and a 3-year outlook.
Most company AI starts with knowledge retrieval. This track builds it properly and privately.
Private RAG
Systems
Builder
Track°
Build a retrieval system on your own documents that answers with sources, respects permissions and runs where your data lives.
Ingestion &
Chunking
Clean, retrievable
knowledge
Turn documents, mail and archives into clean, retrievable knowledge.
Embeddings &
Vector Stores
Where quality
is decided
Choose, host and tune embeddings and indexes locally or in cloud.
Retrieval
Evaluation
Measure every
answer
Measure answer quality, citations and failure cases.
Permissions &
Audit
Right answer,
right person
Role-aware retrieval and logs your compliance team will accept.
Local
Deployment
Nothing leaves
the building
Run the full stack on your own hardware, no data leaves the building.
Every system needs someone accountable for it. The AI Officer course prepares that person.
AI Officer
Training
Governance
Certification°
The practical governance course for the person who answers for AI in the organisation: EU AI Act, risk, literacy, policy and incidents.
EU AI Act
Essentials
Roles and
obligations
Obligations, timelines and roles for providers and deployers.
Risk
Classification
Know your
risk class
Classify your systems and document the reasoning.
AI Literacy
Programme
Article 4,
in practice
Design the staff AI-literacy programme Article 4 expects.
Policies &
Audits
Paper that
matches practice
Usage policy, vendor review, model registry, internal audit.
Incident
Response
When AI
fails
What to do when an AI system fails, leaks or misleads.
Agents widen the attack surface. Security teams need a track of their own.
Cognitive
Security
Defense
Track°
The new attack surface: agents that read mail, browse and call tools. Learn how they are attacked and how to defend them.
Prompt
Injection
Hijacked by
a document
How agents get hijacked through documents, mail and web pages.
Agent Data
Exfiltration
Least-privilege
agents
Tool permissions, egress control and least-privilege agents.
Deepfakes &
Social Engineering
Doubt the
convincing
Detect and respond to synthetic voice, video and text attacks.
Red-Teaming
LLM Apps
Attack yourself
first
Attack your own assistants before someone else does.
AI is moving into the physical world. The last track covers twins, robots and safety.
Digital Twins
& Robotics
Embodied AI
Track°
From ROS 2 basics to simulation-first development and humanoid operations, with safety assessed at every step.
ROS 2
Foundations
Nodes, topics,
services
Nodes, topics and services for teams new to robotics.
Simulation
First
Twin before
robot
Build the twin, test the task, then touch the robot.
Humanoid
Operations
Operator
copilots
Maintenance, knowledge capture and operator copilots.
VLA Models &
Safety
Capability and
safety
Vision-language-action models, limits and safety cases.
District-AI° Private AI Systems
Private AI Systems.
Founder-led. Confidential.
Built for regulated and
creative teams.