Challenges

This hackathon is a tree of three tiers. Tier 1 · Foundations is one guided, linear challenge — four ordered steps everyone completes. Tier 2 · Advanced is a set of modular, self-contained challenges (plus Extras) you can attempt in any order. Tier 3 · Capstone is an open-ended design brief that composes everything into a multi-agent system. Every Advanced challenge and the Capstone assume the Foundations end-state: a deployed, grounded Northfield University IQ Assistant that answers from the FAQ corpus with citations.

  TIER 1  FOUNDATIONS (guided, linear, everyone)
    Step1 --> Step2 --> Step3 --> Step4  <-- Foundations end-state
           |
           v
  TIER 2  ADVANCED (modular, pick any order)
    Action Tools | Evaluation+RedTeam | Tracing | Deploy
    deepeners: Fabric IQ | Voice Live | Build a UI | Copilot-Assisted
           |
           v
  TIER 3  CAPSTONE (open-ended design brief)
    Northfield IQ multi-agent: triage/router fans out to specialists
    (knowledge, actions), then converges.

Tier 1 · Foundations

One guided challenge, four ordered steps. Each step ends in a Checkpoint (python validate.py --step N) that is the prerequisite for the next. Completing Step 4 is the Foundations end-state — the gate to the entire Advanced tier.

Step Focus Outcome
1 — Setup & Provisioning azd up (Foundry + AI Search + App Insights), keyless auth Infrastructure live and authenticated
2 — Model & Playground Deploy + compare models, tune system instructions Generic answers from a model you chose
3 — Your First Agent Named, versioned Foundry agent with persona + guardrails An agent that refuses out-of-scope/unsafe asks
4 — Knowledge Base (IQ) Index the FAQ corpus, build Foundry IQ, attach to the agent Grounded answers with citations — END-STATE

Start Foundations →

Tier 2 · Advanced + Extras

Pick any of these in any order. Each one extends the same assistant and assumes the Foundations end-state (or the bootstrap skip-path below).

Advanced challenge Guided / Scratch What it adds
Action Tools ~45 min / ~1.5 hr The agent does work — opens a ticket / places a hold / books advising via an MCP tool, with a human-approval loop
Evaluation & Red Teaming ~1.25 hr / ~2 hr Proof it’s accurate and safe — groundedness metrics + adversarial/jailbreak results, with a CI score gate
Tracing & Observability ~1 hr / ~1.5 hr Every answer observable end-to-end — OTel GenAI spans → App Insights → KQL
Deploy as a Hosted Agent ~60–90 min / ~1.5 hr Ship it as a containerized hosted agent with its own endpoint and per-agent identity

Each Advanced challenge offers a Guided path (revised, honest time) and a longer Build-from-scratch path — both graded by the same validate.py.

Extras, re-slotted by their role in the tree:

Extra (optional) Tree role What it adds
Magentic Workflows Capstone-feeder Multi-agent orchestration with a Magentic manager (MAF)
Hosted Long-Running Agents Capstone-feeder MAF + hosted long-running agent patterns
Build a UI Capstone companion A web front-end: streaming chat, citations panel, action-approval card
Fabric IQ Deepener Live operational data grounding via OneLake / Fabric IQ
Give It a Voice Deepener A spoken interface with the Voice Live API
Copilot-Assisted Build Deepener Use microsoft/skills + MCP + copilot-instructions.md to build faster

Tier 3 · Capstone

The open-ended summit: break the single Northfield IQ Assistant into a multi-agent team — a triage/router that fans out to specialist agents (knowledge, actions) and converges — orchestrated with the Microsoft Agent Framework (MAF). It’s a design brief, not a placeholder-fill: you decide the org-chart and wire the graph.

Capstone Time Prereqs
Northfield IQ, the Team — Multi-Agent Orchestration 2–2.5 hr core (+1 hr Magentic stretch, +1.5 hr hosted variant) Foundations end-state + Action Tools

Start the Capstone →

Make it your own: the capstone is the best place to reskin — swap Northfield for your domain (insurance, factory ops, retail) and demo your agent team.

How to run this hackathon (Two Paths)

There are two ways in. Both converge on the same Foundations end-state, then fan out into the modular Advanced tier.

  • Path A — Beginner (default, recommended). Complete Foundations as one guided, linear challenge (4 ordered steps), then pick Advanced challenges in any order.

  • Path B — Advanced-skip (for teams who already know Foundry basics). Run one bootstrap (~10–15 min) that materializes the Foundations end-state, verify a single checkpoint, then jump straight to the Advanced tier. You skip the guided learning, not the prerequisites.

  PATH A (Beginner)
  +---------------------------------------------------+
  | FOUNDATIONS (guided, linear, everyone)            |
  | Step1 --> Step2 --> Step3 --> Step4               |
  | Setup    Model    Agent    Knowledge Base         |
  +---------------------------+-----------------------+
                              |
                              v  Foundations end-state:
                                 deployed, grounded Northfield IQ Assistant

  PATH B (Advanced skip)
  [bootstrap: azd up + setup-foundations] --> same end-state
  (~15 min, 1 checkpoint)
                              |
                              v
  ADVANCED (modular, pick any order)
  Action Tools | Evaluation+RedTeam | Tracing | Deploy
  Extras: Fabric IQ | Voice Live | Magentic | Hosted MAF | UI | Copilot

The bootstrap checkpoint (single gate for Path B):

azd up                                    # provision Foundry + AI Search + App Insights (+ ACR for deploy)
./scripts/setup-foundations.sh            # deploy model, create agent, index corpus, build IQ knowledge base
python scripts/validate-foundations.py    # ✅ asserts the Foundations end-state exists

validate-foundations.py must pass green before a Path-B team starts any Advanced challenge. Every Advanced challenge assumes the same end-state, so you materialize it once, verify it once, and the whole Advanced tier is unblocked.

Need to set up your environment first? Start at Getting Started and return here once your toolchain is ready.

Table of contents

WTH AI Hackathon — Built with ❤️ for students and coaches

This site uses Just the Docs, a documentation theme for Jekyll.