Skip to content
ShinyQ HQ: Back to HQ

← Back to the Labs

L3:fabric-forecasting-dashboard · AI Wing · Featured

Demand forecasting on Microsoft Fabric

A national bakery brand's demand-planning dashboard, with resolved capacity throttling and a complete medallion pipeline.

Client
Sari Roti
Role
Sole engineer, responsible for operations
Period
Aug 2026 to Present

Results

  • ~21 s to ~1.1 s

    Time to first byte after addressing throttling

    Measured after removing capacity-heavy indexes and switching to streamed aggregates.

    Verified

  • 61.0 s vs 9.5 s

    Identical 140-request load on small vs larger capacity

    Capacity probe isolated the admission delay per request caused by throttling.

    Verified

  • ~5 weeks

    First commit to production deployment

    Based on repository and deployment history.

    Verified

  • 30 PRs

    Pull requests delivering the medallion pipeline in two days

    Pull requests are merged; live cut-over to the new pipeline remains in progress.

    Verified

Problem

Planners needed a demand-forecasting dashboard on Microsoft Fabric. After a fast production launch, loads stalled for about 21 seconds. The next requirement was a pipeline using the client's own source data.

What I did

  • I deployed a static Next.js app on Fabric, reading lakehouse views through a GraphQL data API and validating results with a 168-check reconciliation harness.
  • I used zero-row and capacity probes to confirm capacity throttling, correcting my earlier diagnosis. I replaced an 83,824-row read with about 3,000 streamed aggregates.
  • I added a private, Entra-protected backend for AI executive briefings. After measuring slow non-streaming responses, I moved chat to a Microsoft 365 Copilot agent.
  • I built one fail-fast, stage-resumable medallion pipeline: data contract, bronze ingestion, silver data-quality rules, forecast adapter, gold serving tables and app sync with reconciliation.

Architecture

  1. ClientNext.js app on Fabric
  2. ServiceData API Builder GraphQL
  3. AIAzure AI Foundry briefingPrivate App Service backend
  4. DataFabric lakehouseBronze, silver, gold
  5. ExternalClient source data

Data flow

  • Next.js app on Fabric to Data API Builder GraphQL
  • Next.js app on Fabric to Azure AI Foundry briefing (executive briefing)
  • Data API Builder GraphQL to Fabric lakehouse (gold views)
  • Client source data sends asynchronously to Fabric lakehouse (Data Pipelines)

Tech stack

  • Microsoft Fabric
  • Fabric Data Pipelines
  • Spark notebooks
  • Lakehouse
  • Data API Builder
  • Next.js
  • App Service
  • Bicep
  • Key Vault
  • Azure AI Foundry
  • Microsoft 365 Copilot

L2 · Related career entryTechnical Consultant, Software & AI · Metrodata (PT Mitra Integrasi Informatika)