All work
AI / LLM Engineering

AI Adoption Analytics Dashboard

A full serverless analytics platform giving IT leadership real-time visibility into enterprise AI tool adoption, replacing manual, ad hoc reporting entirely.

100%

Manual monthly reporting eliminated

Daily ingestion, monthly AI reports

Pipeline cadence

Fully serverless (Lambda + DynamoDB)

Architecture

The Problem

Leadership wanted to understand how a company-wide AI tool subscription was actually being used — adoption over time, power users, trends — but the only way to answer that was manually exporting data and building reports by hand every month.

My Approach

  • Built a serverless pipeline that runs on a daily schedule, pulls usage data from the AI provider's usage API, cross-references department information from the identity/directory API, and persists historical snapshots to a managed NoSQL database.
  • Built a public-facing but access-gated web dashboard (charting library, vanilla JS/HTML) served from static hosting behind a CDN, calling a serverless API layer for live and historical queries — including date-range aggregation across arbitrary historical windows.
  • Automated AI-generated executive board reports: on a monthly cadence, the system uses a generative model to synthesize the month's usage data into a written summary for leadership, removing a recurring manual writing task.
  • Iterated the system across multiple real releases — added manual backfill support for data gaps, fixed a silent pagination bug that had been truncating large historical scans, and continuously tracked technical debt in a living improvements backlog rather than letting it go untracked.
  • Locked down access to the dashboard at the edge (CDN-level request validation) plus an API key, appropriate for an internal analytics tool.

Stack

AWS

LambdaAPI Gateway (HTTP API)DynamoDBEventBridge (scheduled triggers)S3 + CloudFrontIAM

Frontend

Vanilla JavaScriptChart.jsHTML/CSS

Integrations

Anthropic Usage APIMicrosoft Graph API (directory/department data)Gemini API (report generation)

IaC/Deploy

AWS CloudFormation/SAM-style templatesShell-based deploy scripting

Practices

Scheduled data pipelinesHistorical data aggregationLiving tech-debt backlog management

Skills Demonstrated

  • ▸Full serverless application development (API, data store, scheduler, frontend) end to end
  • ▸Data engineering: reliable scheduled ingestion, pagination correctness, historical aggregation
  • ▸Practical use of generative AI to automate a real recurring business task (executive reporting)
  • ▸Engineering maturity: tracking and prioritizing technical debt rather than treating "done" as "shipped once"