Skip to content
at_.DEVELOPMENT
.ABOUT — WHO I AM AND WHAT I DO

Alexander
Talaat_

I build the AI systems a business actually runs on, and I build them to keep running. That means the infrastructure that serves an AI model to real users — not just a demo that works once — plus the tools that let it act on your data and your existing software safely. I also design the interfaces people use to work with it, so the system shows you what it's actually doing instead of leaving you to guess. Fifteen years shipping production web platforms backs that up: I've run systems at real scale long enough to know what breaks, and I build the rollback plan before you need it.

I work mostly with teams who already have something running and need it to hold up — real traffic, real editors, a model that has to answer correctly when things get busy. I like the unglamorous parts: moving data without losing any of it, deploy pipelines that don't wake anyone up at 2 a.m., tests that check whether a system's answers can be trusted, and the moment it stops being the reason people are stuck in a meeting.

PORTRAIT
Taking Q4 projects
Washington, D.C. · remote-first. Fixed-scope project work, engagements capped at 4–6 weeks. No retainers. $150/hr.
15+
YEARS IN PRODUCTION
26
AI MODELS BENCHMARKED
6
CASE STUDIES SHIPPED
D.C.
BASED IN D.C., WORKS REMOTE

Experience

Fifteen years across agencies, newsrooms and product teams. Roles listed with what I was actually accountable for.

2022 — NOW
INDEPENDENT
Principal engineer — AI platforms & architecture
talaat.dev — Washington, D.C.

Design and build the infrastructure that runs AI models in production for newsroom and SaaS clients — the systems that find and hand a model the right information, and the deployment tooling that ships updates without breaking anything. Most recently: hal0, a home AI platform I built solo that runs ten different AI capabilities — chat, search, transcription and more — behind one interface, on hardware I own instead of renting from a cloud AI vendor. Sole engineer on most engagements; technical lead on the rest.

PODMANNEXT.JSNODE.JS
2017 — 2022
AGENCY
Lead full-stack engineer — platform engineering at scale
Digital agency — enterprise WordPress practice

Set the technical standard used to bring forty client websites onto one shared system: a single build pipeline, automated tests run before every release, and a release process that could switch traffic to a new version and instantly roll it back if something broke — no downtime either way. Also led the review process that got drag-and-drop page-builder plugins out of the stack for good, after they kept breaking under real traffic.

WORDPRESSPHP 8CI/CD
2011 — 2017
IN-HOUSE
Web developer → senior developer
Membership organization — D.C.

Rebuilt the organization's public website and member portal, and moved a decade of content off an aging system without losing any of it. Learned firsthand what a content migration costs when nobody plans the fallback in advance — the same discipline that later became the automatic rollback built into the release process above.

Skills

GROUPED BY WHAT I'M HIRED FOR
AI INFRASTRUCTURE
Podman · systemdROCm · llama.cppMCP serversEvals · pgvector
DESIGN & DESIGN SYSTEMS
Figma → CSS tokensAstro · Tailwind 4Accessibility · motionInformation design
APPLICATION ENGINEERING
TypeScript · ReactNext.js App RouterNode.js · tRPCPostgreSQL · Prisma
PLATFORM & DEPLOYMENT
Docker · GH ActionsBlue-green deploysVercel · CloudflareWordPress at scale
OPEN SOURCE
WRITING
2026
Building hal0, part 3: the silence problem
Article · 10 min
2026
Building hal0, part 1: fifteen tags in ten days
Article · 8 min
2026
Evals are just tests with taste
Article · 6 min

Let's Build Something Cool._

Ready to bring your project online? Tell me what you're building.