PythonDjangoLangChainClaudeAzure AI FoundryMicrosoft DataverseEntra IDOAuth SSOJavaScriptSVGPydanticIIS
Overview
A production web application that replaced a spreadsheet based process for tracking Risks, Actions, Issues, and Decisions across enterprise project delivery. It serves project managers working on more than ten concurrent client projects and over eleven hundred live records, and it layers a natural language AI assistant on top of live enterprise data.
The Problem
Project managers tracked risks, actions, issues, and decisions in spreadsheets that were hard to keep consistent, offered no audit history, and could not answer questions across projects. Reporting was manual and error prone, and there was no secure single source of truth tied to the corporate identity system.
Architecture & Key Work
- ▹Designed, built, and deployed a production Django application replacing the spreadsheet based RAID process for more than ten concurrent client projects and over eleven hundred live records
- ▹Integrated Microsoft Entra ID single sign on with django allauth and a Dataverse Web API backend using app only OAuth, giving secure credential free login inherited from the corporate tenant
- ▹Engineered an agentic AI assistant with LangChain and Claude on Azure AI Foundry, using tool calling over live Dataverse data and Pydantic validated structured outputs
- ▹Built a human in the loop draft a ticket flow enabling natural language and voice queries against project data
- ▹Built a client side cross filtering analytics dashboard in vanilla JavaScript and SVG with server side pagination and caching, scaling cleanly to fifty thousand records without a database migration
- ▹Implemented a full audit trail, automated overdue reminder digests, and filter aware PDF and Excel export
- ▹Designed a deterministic project code system guaranteeing unique collision free ticket identifiers
Results & Impact
- ✓Replaced a fragile spreadsheet workflow with a secure single source of truth across ten plus client projects
- ✓Enabled natural language and voice access to live project data through the AI assistant
- ✓Led a build versus buy evaluation against Microsoft Copilot Studio, documenting authentication, governance, and live data limitations to inform enterprise AI tooling strategy
- ✓Owned end to end delivery on Windows and IIS including deployment, configuration, and troubleshooting