-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Welcome to the erate-workbench-poc wiki!
A proof-of-concept prototype analytics platform built on real USAC E-Rate open data — and a full demonstration of what disciplined, AI-assisted software engineering looks like in practice.
ERATE Workbench is a .NET 8 / ASP.NET Core analytics platform that ingests federal E-Rate program data across multiple USAC open datasets (FY2020–present) and surfaces funding analytics, applicant risk indicators, competitive intelligence, and program workflow reference tools — all built against real data, with a documented REST API, automated CI/CD pipeline, and full DevSecOps coverage.
This is not a tutorial project or a throwaway prototype. Every decision made here reflects real-world engineering discipline — from layered architecture and ADRs to secrets scanning, dependency governance, and AI-assisted development workflows designed to be repeatable and governable across any project.
The repository is public at github.com/steven-dracker/erate-workbench-poc.
| Layer | Technology |
|---|---|
| Language | C# / .NET 8 |
| Framework | ASP.NET Core Razor Pages + Minimal API |
| ORM / Database | Entity Framework Core / SQLite |
| API Documentation | Swagger / Swashbuckle |
| Analytics | Chart.js |
| Testing | xUnit / Playwright UI smoke tests |
| CI/CD | GitHub Actions — multi-stage pipeline with artifact publishing |
| Security Scanning | Gitleaks, Semgrep, Dependabot, SonarCloud |
| Data Source | USAC E-Rate Open Data via Socrata API |
| Runtime | WSL2 / Ubuntu on Windows |
The core project documentation. Covers the engineering thesis, architecture decisions, three-layer data model, DevSecOps pipeline, observability approach, and what this project demonstrates across software architecture, full-stack development, quality engineering, and product thinking. Start here for the technical overview.
The multi-AI workflow system used to build this application. Documents how ChatGPT acts as architect and Claude Code acts as implementation engine, with the repository serving as shared memory between both systems. Covers the Boot Block context preservation primitive, slash command conventions (/handoff, /remembernow, /new-task), the CLAUDE.md auto-load system, and how architectural discipline is maintained across stateless AI sessions.
The specification layer that makes AI-assisted development repeatable and governable. Covers how stakeholder requirements — functional, non-functional, and architectural — are structured into prompts that produce consistent, scope-controlled code. Includes the three-level specification hierarchy (Project Constitution → Feature Specification → Task Prompt), schema discovery as a mandatory prerequisite, the context transfer protocol, and the prompt traceability convention that creates an audit trail from backlog item to committed code.
A strategic overview of Jira's full project management capability beyond ticketing — covering agile sprint-based delivery and waterfall milestone-driven projects, the work hierarchy from initiatives to subtasks, velocity and predictability, cross-team portfolio management, and applicability to non-technical business functions including marketing, HR, finance, and client services. Includes a comparison of alternative platforms — Monday.com, Asana, and Smartsheet — evaluated through the lens of a consulting and services business.
| Discipline | Evidence |
|---|---|
| Software Architecture | Layered .NET solution, ADRs, three-layer data model |
| Full-Stack Development | ASP.NET Core, Razor Pages, Minimal API, EF Core, SQLite |
| DevOps | Multi-stage GitHub Actions CI pipeline, artifact publishing |
| DevSecOps | Gitleaks secrets scanning, Semgrep SAST, Dependabot, SonarCloud |
| Quality Engineering | xUnit unit tests, Playwright UI smoke tests, reconciliation validation |
| AI-First Development | Multi-agent workflow, boot block context system, prompt traceability |
| Domain Knowledge | E-Rate program lifecycle, USAC open data, funding analytics |
| Product Thinking | Real domain thesis, stakeholder-driven feature design |
Steven Dracker is a technology leader with 30 years of experience in software development, IT, and cybersecurity — primarily in leadership and architecture roles. He has built and led distributed engineering teams, holds deep expertise in compliance-driven technology environments including PCI Level 1 and SOC 2 Type II, and is an active practitioner of AI-assisted software development.