Support Kit
Dynamics 365 F&O
A public diagnostic tool for Dynamics 365 Finance and Operations. It surfaces the current form and table name on any page, searches across more than 120 tables, builds deep links straight to specific records and forms, and captures a full context snapshot, screenshot included, so an issue can be reported precisely instead of described from memory. Version 1.5.0 is live on the Chrome Web Store, rebuilt onto Microsoft’s documented navigation API after an earlier approach using undocumented URL parameters failed in live testing.
Freight alert
Automation
A shipment once missed its booking window because a flagged change sat unopened in a portal that only got checked once a day. That gap became a Copilot Studio agent and Power Automate flow that reads incoming shipment data as it arrives, checks received quantities against what was ordered, and posts any discrepancy straight to Teams the moment it appears. What used to depend on someone remembering to log in now happens on its own, with a written trail kept automatically.
Self-serve
Buffer Reset Tool
A recurring process where hitting a capacity limit on a data export table meant waiting for someone to manually clear it through a database tool became a Power App that authorised users can trigger themselves, with every reset logged for audit. I added a live capacity gauge on top so the table filling up is visible before it becomes a problem, rather than found out after it stalls something else.
Diagnosing pipeline
Server Migration
Diagnosing a broken pipeline after a server migration After a server migration, an order intake pipeline quietly stopped working. The visible failure looked like a data problem, which is where the first investigation went. It turned out to be a protocol mismatch two steps upstream, an automation flow still using a file transfer method the new server no longer accepted, failing silently on every run. Tracing a failure back past the system where it shows up is usually more useful than the specific fix itself.
AI Agent
WhatsApp Sales Agent
Built a WhatsApp based sales agent using Node.js and the Claude API, designed multi-tenant from the start on Postgres and backed by a queue and cache layer for reliable message handling. The channel logic sits behind an adapter pattern deliberately, so adding voice as a second channel later will not mean rebuilding the conversation logic underneath it.
Churn prediction
Freight & Logistics
Built a churn prediction model for freight and logistics customer data using XGBoost, reaching a 0.935 ROC AUC score. Rather than stopping at the prediction, I added SHAP based explainability so the reasoning behind each flagged account is visible rather than just the flag itself, and served the model through a FastAPI backend so it could sit behind an actual application instead of staying in a notebook.
Churn prediction
Telecom
Built a telecom churn prediction model using a Random Forest classifier, reaching a 0.984 ROC AUC score, paired with a browser based tool that lets a user query the results through Claude, GPT-4o, Gemini or Groq interchangeably, so the choice of model provider is not locked in at build time.
monitoring dashboard
Shop Floor
Built a four tab offline dashboard for shop floor monitoring: live OEE gauges, a Monte Carlo simulation for output forecasting, predictive maintenance alerts, and a scenario planner for testing changes before committing to them on the actual floor. It runs entirely offline, so exploring how it behaves never depends on a live data connection.
Have Project In Mind?
Let's Talk!
Interested in working together? Get in touch and let’s talk through what you need.
My Expertise
- Microsoft Dynamics 365
- Power Platform
- Azure
- SQL Server
- Python
My Skills Set
- AI & Automation
- Data Engineering
- Data Analytics & Machine Learning
- Software & Tool Development
Contact Details
- United Kingdom
- support@khadijaahmad.com
- 123-456-7890