AI Platform Architect and Systems Builder.
I build software that has to work outside the demo: local and self-hosted AI platforms, inference systems, governed multi-agent engineering workflows, developer platforms, enterprise systems, and durable web architecture. This repository is a map of that work.
Senior software engineering is the foundation of my work. I’m based in New York City, and an Executive MBA plus years of product and project experience help me weigh the business case alongside the technical one. The aim is straightforward: clear boundaries, useful systems, and results people can rely on.
-
AI Back Office (presentation, case study)
A connected back office bringing Google Workspace, Gemini, department workflows, and an executive dashboard into one operating view. -
Fernain
Connected software for membership organizations and modern teams, with shared identity, permissions, and governance across focused products. -
Memoriosa
A private, local-first home for memories, connecting people, moments, and stories so you can return to them with their context intact.
-
I Built an AI Back Office (slides, 9 pages)
The first publication card has two tabs: Slides downloads the presentation PDF; the Case study PDF downloads the 16-page build guide, Building an AI back office with Google Workspace. -
AI-Assisted Software Delivery Management (research note, v1.1.0, 57 pages)
Manage AI work without pretending agent activity equals delivery. A practical AI Delivery Execution Profile for human ownership, attempt lineage, evidence-bound acceptance, separate clocks, and cost per accepted outcome. -
Provider-Agnostic AI Orchestration, Volume 1 (research note, v1.3.0, 26 pages)
Route work by role, evidence, and cost - not provider loyalty. A candidate framework for provider-agnostic roles and contracts across cloud and local execution, with human authority and evidence-based acceptance gates. It is not a provider benchmark. -
AI-Assisted Estimation Methodology (methodology, v1.2.0, 20 pages)
Estimate delivery without confusing AI speed with human capacity or elapsed schedule. A practical method for separating baseline effort, operator load, AI execution, validation, and schedule. -
On the site, Provider-Agnostic AI Orchestration Volume 1 is the only active volume tab; Volumes 2-6 remain disabled and are not published.
-
Local Codex for Enterprise
A self-hosted AI development platform for teams that need local inference, clear access boundaries, and an auditable trail. -
TootyCMS
A modular content-management platform focused on extensibility, reusable architecture, and long-term maintainability. -
Codex for Docker
Local-first developer tooling built around Docker Model Runner and OpenAI-compatible endpoints. -
Rust MCP Filesystem Parsers
Rust-based filesystem and document parsing for MCP integrations and AI retrieval workflows.
- Systems credited with more than $3M in organizational savings
- Helped scale a digital job platform from approximately $500K to $1.5M
- Replaced a projected $60K GIS implementation with an approximately $4K solution
- Built internal platforms that avoided roughly $180K a year in external costs
- Start with the problem. Know the boundary before writing code.
- Use AI as a tool, where applicable. It can write code. I still own the design and result.
- Keep the system legible. The next engineer should be able to follow it.
- Test everything I build. TDD first where possible. Working code still needs proof.
- Ship work that lasts. It should hold up after the demo.
This repository contains the static source for my personal site. See UPDATING.md before changing project order, public status language, links, or the wider story.
Find the code at GitHub, connect on LinkedIn.
For the full visual story, current project highlights, and more about how I work, visit joseffb.com.