Microsoft data engineering for ERP reporting, Fabric and AI agents.
Meles helps companies turn scattered operational, financial and ERP data into trusted reporting, automated data flows and practical AI/data agents.
Microsoft Fabric · Dynamics 365 F&O · Power BI · Azure
Most companies don't have a data problem. They have five small ones.
Individually they look manageable. Together they mean nobody fully trusts the numbers — and every report needs a caveat.
- →Two reports show two different revenue figures, and nobody can say quickly which one matches the ledger.
- →Month-end reporting depends on someone exporting from the ERP and rebuilding the same Excel file, every month.
- →Systems are connected through integrations that only one person understands — and that person is busy.
- →Power BI reports exist, but each one queries the sources in its own way, with its own definitions.
- →There's interest in AI on company data, but no safe, governed way to let it near finance systems.
Three kinds of work, one foundation
Meles is a Microsoft-stack data engineering practice. The work centres on Fabric, Dynamics 365 Finance & Operations, Power BI and Azure — and on making finance and operational data dependable enough to build on.
Reliable data foundations
One governed place where ERP, operational and financial data lands, gets cleaned, and gets defined once — typically a Fabric lakehouse with documented pipelines and a shared semantic model.
Reporting people trust
Power BI reporting built on that foundation, reconciled against the source systems, so finance and operations look at the same figures — and know where they come from.
Practical data agents
Scoped, read-only agents that answer business questions against the governed data — with documented assumptions and protected access to sensitive finance data. Not a generic chatbot.
Meet Albert, the Meles data agent — read two sample reports he produced, or use the live AR Control Tower app built under the same rules.
The same five steps, every engagement
No black boxes. Each step produces something you keep: documentation, working code, or a tested deliverable in your own environment.
-
Analyse the existing data sources
What lives where, how reliable it is, who owns it, and what the reports actually need. This step usually surfaces the real problem — often it isn't the one in the original request.
-
Design the data model
A documented model that defines each business concept once — revenue, margin, open orders — so every report and every agent works from the same definitions.
-
Build pipelines, notebooks or API integrations
Automated flows that move and transform the data: Fabric pipelines, notebooks, and API integrations between the systems that need to talk to each other.
-
Add logging, error handling and monitoring
Pipelines fail sometimes; the difference is whether anyone notices. Every build ships with logging, alerts and clear failure behaviour, so problems surface before the numbers do.
-
Deliver in Power BI, Fabric or Azure
The result lands in your environment, under your access controls, with documentation and a handover — not a dependency on Meles.
Built for the people who live with the numbers
Business owners & operations leaders
You want to see how the business is actually running — orders, stock, margin, cash — without waiting for a spreadsheet or wondering whether it's right.
Finance leaders
You need reporting that reconciles to the ledger, closes faster, and doesn't depend on fragile manual exports from Dynamics 365.
IT & data teams
You want a clean architecture on Fabric, documented pipelines and integrations, and a sane governance story for reporting and AI access.
Companies with scattered ERP data
Multiple systems, unclear ownership, reporting held together by exports — and a sense that the data could be doing much more.
One engineer, end to end
Meles is the practice of Tim, a Belgium-based data engineer specialised in Microsoft Fabric and Dynamics 365 Finance & Operations. The person who analyses your sources is the person who designs the model, writes the pipelines and reconciles the output against your ledger — no handovers between sales, architects and builders.
That background sits deliberately on both sides of the fence: the D365 source reality — finance, procurement, inventory and logistics processes, and the tables underneath them — and the Fabric, Power BI and Azure platform that has to report on it. When a number doesn't add up, it gets traced end to end: business process, ERP entities, lakehouse layers, transformation logic, API payloads, report.
Working principles, not promises
Meles doesn't publish invented case studies or borrowed logos. What it can show you is exactly how the work is done — and these principles apply to every engagement.
Read-only by default
Production finance systems are accessed read-only unless a change is explicitly scoped and approved. No exceptions for convenience.
Documented assumptions
Every mapping, filter and definition is written down. If a report says "revenue", the documentation says exactly which tables, which statuses, which exclusions.
Your environment, your keys
Everything is built in your tenant, under your access controls. Handover is part of the deliverable, not an extra.
Sensitive data stays protected
Ledger, vendor, customer, invoice and payment data is handled through scoped queries, metadata checks and sanitized outputs — especially where agents are involved.
Start with one report you don't trust
The most useful first conversation is concrete: one report, one integration, one recurring Excel exercise. Describe it, and you'll get an honest view of what fixing it would take.