recall&build

LEARN BY BUILDING / AGENT ENGINEERING

Build an agent.
Know how it works.

Choose your technology. Build a small project. Explain the decisions with working code and evidence.

01 Recall the concepts02 Build on your laptop03 Test what happened

Short lessons for your commute.
Hands-on work for your next free hour.

I WANT TO BUILD AN AGENT

Choose your platform.

MicrosoftAgent Framework

Build a support agent with a reusable skill, scoped tools and approval boundaries.

2–3 hours1 project10 recall cards

You’ll leave withA support-agent project, a skill bundle and a tested evidence pack.

Start the Microsoft pathReview the Microsoft cardsOfficial documentation

Time estimates cover a first pass, not mastery. Core exercises run locally; optional API runs use your own account.

YOUR BUILD SEQUENCE

01
AGENT FRAMEWORK · 2–3 hours

Build a support agent with a reusable skill

Use skills, scoped tools, approval and evidence in one agent project.

Notebook, starter code and tests
Open labNeed the foundations first?Four short Azure AI videos, with citations and companion labs.
Every path: source links · runnable exercises · saved checkpointsBrowse all projects

LEARN / MEASURE / CHECK

Understand inference engineering.

Ten readable cards cover serving, memory, scheduling and GPU tradeoffs. Work through examples, then check your understanding with saved questions.

Open inference cards →

COMMUTE SESSION / AZURE ARCHITECTURE

See how the pieces fit.

A visual map and sixteen short cards explain Microsoft’s AI architecture. Follow a request, explore context and tools, then check the operating tradeoffs.

Open visual flashcards →

DEEPER BUILD / CODING AGENTS

Build the harness from scratch.

Follow the eight-lesson Decode course: model loop, permissions, sandboxing, context, subagents, evaluations and replay. Source-linked lessons, spec-driven practice, recall cards and saved project checkpoints.

Explore the coding-agent path →

FROM BUILDING TO EXPLAINING

Defend your engineering decisions.

Ten interview drills, linked to your labs. Diagnose a retrieval failure, stop an agent loop and explain a measured improvement.

Open interview practice →

LEARN FROM WHAT PEOPLE BUILD

Expert Spotlights

Builders across the learning paths.
Real projects, with the source to inspect.

OpenAI · AGENTS API

Rach Pradhan

Folio team, with Yu Xi Lim

Best use of Agents API

Singapore · 13 September 2026

THE PROJECT / FOLIO

Inspect the answer.
Follow the evidence.

Rach Pradhan and Yu Xi Lim built Folio, a tool for inspecting how AI answers recommend websites and cite sources. Organizer Gabriel Chua named their pair among the Best use of Agents API winners at the GPT-6 Astra Hackathon in Singapore. Explore the project’s public submission and source code, then examine its approach to saved evidence and repeatable comparisons.

Explore FolioPublic repositoryAward announcement

Award verified from the organizer announcement; project capabilities described by its creators and public repository. An editorial spotlight; no teaching session or endorsement is implied.