Developers, technical analysts, and engineers moving into applied AI.
PROGRAM 01 / BUILD RELIABLE AI APPLICATIONS
Applied AI Engineering
Turn a language model into a useful application. Build agents, connect private knowledge, and learn how to test the whole system.
- Python & Pydantic
- Agents & MCP
- Private knowledge
- Evaluation & deployment
Working knowledge of Python. You should be comfortable reading code and using a terminal.
A private knowledge assistant with source references, approved tools, evaluation cases, and a deployment guide.
THE CURRICULUM
From foundations
to a working project.
Open a module to see its agenda and practical outcome. The learning path builds across three stages.
STAGE 01
Engineering foundations
01The AI application, end to end
- Map the user task, inputs, decisions, and acceptance criteria.
- Set up a reproducible Python project with uv and version control.
- Use a coding assistant with clear project instructions and review checks.
02Python for dependable services
- Use type hints, async operations, and explicit error handling.
- Separate application logic from providers and infrastructure.
- Test retries, timeouts, and failure recovery.
03Structured data with Pydantic
- Define inputs, tool arguments, and model outputs as schemas.
- Validate incomplete or unexpected responses.
- Keep parsing errors visible and recovery bounded.
04Models, prompts, and provider APIs
- Compare hosted and local models against a real task.
- Work with structured output, tools, and document inputs.
- Measure response quality, latency, and token cost.
STAGE 02
Knowledge & agent workflows
05Build a tool-using agent
- Create a typed agent with PydanticAI.
- Connect narrow tools with explicit permissions.
- Control tool errors, retries, and stop conditions.
06Private document retrieval
- Prepare documents, chunks, metadata, and access boundaries.
- Store embeddings in Qdrant and combine retrieval methods.
- Return source references and evaluate retrieval failures.
07State, handoffs, and human review
- Model a workflow as explicit states and transitions.
- Preserve useful context without retaining unnecessary data.
- Add review points before consequential actions.
08Connect systems through MCP
- Understand clients, servers, tools, and resources.
- Expose a small set of approved business operations.
- Test authentication, permissions, and unsafe tool requests.
STAGE 03
Production & final project
09Evaluation and observability
- Build a case set from expected inputs and difficult failures.
- Trace model calls and tool decisions with Langfuse.
- Compare prompt and model changes against the baseline.
10Serve and deploy the application
- Expose the workflow through FastAPI.
- Package services with Docker and manage runtime configuration.
- Add health checks, budgets, and recovery instructions.
11Security and reliability workshop
- Test untrusted documents and prompt injection attempts.
- Apply least privilege and protect application secrets.
- Exercise unavailable providers and misleading model output.
12Final project: private knowledge assistant
- Combine retrieval, typed tools, and a reviewable interface.
- Demonstrate the workflow against the agreed evaluation cases.
- Present limitations, operating cost, and the handover guide.
AI ENGINEERING / WAITLIST
Your next step
starts here.
Join the waitlist for Applied AI Engineering. We will send the cohort dates, delivery schedule, and enrolment details when they are ready.
AED 4,999 full program fee
This waitlist is only for this program. You can join another program separately.
EXPLORE ANOTHER DIRECTION