UNIVO School of AIFind your programme
2026 curriculum preview · Proposed live programmes; cohort dates, fees and teaching assignments are being finalised.
All programmes / Technology

LIVE ONLINE · PYTHON REQUIRED

Agentic AI & Autonomous Systems

Build agents that use tools, handle failures and earn trust through evaluation.

12 weeks48 live hoursCapstone + live defence

YOUR LEARNING JOURNEY

From understanding
to application.

Curriculum 2026.09Reviewed 2026-09-07
01
WEEKS 1–2 · 8 LIVE HOURS

From LLM calls to dependable agents

  • Structured outputs and tool calling
  • Workflows versus autonomous agents
  • State and step budgets
  • Deterministic baselines
Your practical output

Build a single-agent workflow and compare it with a simple baseline.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.
02
WEEKS 3–4 · 8 LIVE HOURS

Context, memory and grounded knowledge

  • Context engineering and compression
  • Retrieval with provenance
  • Short-term state versus durable memory
  • Memory consent and isolation
Your practical output

Create a grounded assistant with source citations and memory controls.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.
03
WEEKS 5–6 · 8 LIVE HOURS

Tools and interoperability

  • API schemas and authentication boundaries
  • MCP clients and servers
  • Least-privilege tool design
  • Idempotency, retries and approval gates
Your practical output

Connect a read-only tool and a separately approved write operation.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.
04
WEEKS 7–8 · 8 LIVE HOURS

Orchestration and human collaboration

  • LangGraph-style stateful graphs
  • Agent handoffs and specialist roles
  • Durable execution and recovery
  • When multi-agent overhead is justified
Your practical output

Implement a recoverable workflow with a human escalation point.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.
05
WEEKS 9–10 · 8 LIVE HOURS

Evaluation, security and operations

  • Task success and tool-call correctness
  • Trace-based error analysis
  • Prompt injection and excessive agency
  • Cost, latency and observability
Your practical output

Build an evaluation set and test adversarial and failure cases.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.
06
WEEKS 11–12 · 8 LIVE HOURS

Deployment studio and live defence

  • Versioning prompts and tools
  • Sandboxed execution and secret management
  • Canary releases and rollback
  • Runbooks and maintenance
Your practical output

Deploy the capstone and defend quality, safety and cost evidence.

Submit the working artefact and a short evidence note; discuss feedback in the live clinic.