NavAIgate Build Challenge 01

The Insight Engine

Build the product pattern enterprises pay six figures for: a chat-first analytics engine that turns a raw data warehouse into insights, decks and decisions. Best build wins a Claude Max 20x plan, on us, for one month.

Any stackAny cloudThree weeksUltra polished or go home

Read the brief Get the dataset How to enter

The brief

Eight points. All eight working is the bar for a serious entry. Build them in the order that suits your stack — but ship them all.

Data in

Connect to the official Aurelia Cloud dataset (a Snowflake-style warehouse extract we host) and load it into the engine of your choice.

Chat-first

A conversational interface is the primary way users explore the data — ask questions in plain English, get answers grounded in the warehouse.

Graphs and sales intelligence

Generate charts and surface commercial insight — pipeline, churn risk, whitespace, adoption trends. There are real stories hidden in the data; find them.

Deck generation

Produce a presentation from any analysis — an in-browser HTML deck and a downloadable PowerPoint (.pptx), both rendered by your application at runtime. No Claude Code, no coding assistant, no external pipeline doing it off-stage: judges will generate decks live inside your app.

Login

SSO (Google / Microsoft / GitHub) or email auth. Every user sees only their own work.

The gallery

A space where rendered decks and artefacts live after the conversation — each one linked back to the chat and data that produced it. The app remembers.

Open the engine

Expose an MCP server or a documented API (either) so agents can drive your product, not just humans.

Ship it

Live deployment on Railway, AWS, Azure, Vercel, Cloudflare — anywhere. Judged as a product, not a prototype.

The dataset — Aurelia Cloud

A fictional B2B design-software SaaS. Eight tables, roughly 620k rows, eighteen months of usage — fully synthetic. There are real commercial stories buried in here. Your engine should find them.

Table inventory

accounts500

contacts2.5k

products12

subscriptions1.8k

usage_daily600k

support_tickets8k

opportunities1.2k

nps_responses3k

Downloads

Full dataset

CSVs + SQLite + DDL, zipped.

Download .zip

Data dictionary

Every table and column, explained.

Open .md

Schema

The DDL — tables, keys, types.

Open .sql

Sample queries

Starter SQL to get you moving.

Open .sql

Synthetic data only. The Aurelia dataset is fictional — do not enrich it with real company data.

What it could look like

Five sketches, five different readings of the same brief. Yours should look like none of them.

A dark analytics workspace with an AI chat panel on the left and live charts and churn-risk account cards on the right. — Mission control — a dense, dark workspace.

A clean, light, single-column chat interface with charts rendered inline and a prominent Generate deck button. — Chat-first minimal — one calm conversation.

A grid of rendered deck thumbnails, each card showing its lineage back to a conversation and a query. — The gallery — every deck remembers where it came from.

A three-pane layout: a warehouse schema browser on the left, chat in the centre, and a live deck preview on the right. — The explorer — schema, chat and deck side by side.

An agent console with a streaming session log on the left and artefacts appearing in a tray on the right as the agent works. — Agent console — driven by an API, watched live.

Rules and eligibility

Read these before you start. Entry constitutes acceptance.

Open to all members of the NavAIgate Skool community — trial members included. Solo or pairs.
Any stack, any cloud, any models. Your API spend is yours.
Must be live and accessible to judges (provide test credentials).
Decks must be rendered in-app, at runtime, by the product you built. Anything generated outside the app — Claude Code, a coding assistant, a manual export — scores zero on that spec point.
You own your work. By entering, you grant NavAIgate a worldwide, perpetual, royalty-free licence to review your submission, showcase it, and adapt concepts and approaches from it in NavAIgate’s own products and client work. Entry constitutes acceptance.
Synthetic data only — the Aurelia dataset is fictional; do not enrich it with real company data.

Judging

Scored against a fixed rubric. Judged personally by Daniel Wright; top entries get a live teardown on Skool.

Spec coverageAll eight points working

40%

Polish and UXDoes it feel like a product

20%

Architecture and write-upClear README + diagram

15%

Insight qualityWhat you actually found in the data

15%

Wow factorThe thing we did not expect

10%

The prize

Built to be worth winning.

Winner

Claude Max 20x plan

Paid by NavAIgate for one month. The serious tier, on us.

Runners-up

Featured on navaigate.dev

Plus a 1:1 architecture session with Daniel.

Timeline

Three weeks. No extensions.

Now

Entries open.

Next community call

Questions? Bring them to the next community call — live Q&A on the brief and the dataset.

Saturday 4 July 2026, 23:59 CET

Deadline. Final submissions in.

Friday 10 July 2026

Winner announced live on Skool.

How to submit

Post in the Build Challenge thread in the Skool community. Include everything below.

A live URL.
Your Git repository — mandatory. No repo, no entry. Private is fine; invite the judge.
A 3–5 minute demo video.
Judge login credentials.
Your README.

Or emaildw@navaigate.dev, subject “Build Challenge 01”.

FAQ

The questions that come up most.

Can I use no-code tools?

Yes, if all eight spec points work — the MCP/API requirement means you will need some backend.

Can I reuse patterns from the classroom?

Yes — encouraged.

Who owns the IP?

You do; see the rules for the licence you grant by entering.

Does it need to handle other datasets?

No — Aurelia only, but generic engines impress.

How big can my team be?

Two, maximum.