The Context

TalkingPoints expanded from messaging into attendance intervention—a new product vertical. Not an incremental feature. A new product that needed new data infrastructure.

I designed and built the D.R.E.A.M. API (Data Rules Everything Around Me 👐🏻) from the ground up.

The Evolution

It started with student attendance tables. Then tardy tracking. Then filtering by absence reason. Then customizable intervention cards that educators could target at specific attendance patterns. Each feature built on the last—the API grew with the product.

Rostering Infrastructure

For rostering data, integrated RosterStream for real-time student rosters—then helped migrate to a Kinesis Firehose architecture when we needed better streaming capabilities.

The data had to be fresh. Attendance products don't work with stale rosters.

The Architecture Decision

Designed it with modern tools: Pydantic for validation, FastAPI for the framework.

This gave us parallelization, fast serialization, and proper validation out of the box.

It also meant that when AI agents became a thing, we could develop new features much faster—the architecture was already set up for it. Type hints everywhere. Schema validation automatic. AI-assisted development worked well because the codebase was structured for it.

API Features

• Student attendance tables with configurable date ranges
• Tardy tracking aggregated at student and school level
• Absence reason filtering for targeted intervention
• Customizable intervention cards with AI-powered targeting
• Real-time rostering via Kinesis Firehose

Impact

• New product vertical enabled from data layer up
• Back-to-School ready with load testing validation
• AI-development friendly architecture with full type coverage
• Real-time data freshness for attendance features

Technologies