AIEngineer&Full-StackDeveloper.
I'm Amin. I started by designing and building websites, then moved into full-stack and cloud development as the projects became more technical. Today, I focus on AI-native applications, working across the backend, data flows, AI workflows, and the interfaces people use.
Amin Hashemi
AI Engineer
Real products, shipped to real users.
Where it started: earlier builds.
A few earlier projects from my WordPress days, and more along the way.
Forpac
Offlineforpac.se · 2025
As an intern developer, I helped build the new Forpac.se website on WordPress with Elementor and the Hello theme for the cleanest possible foundation. The new site is still under construction.
White Tree Turkey
Offlinewhitetreeturkey.com · 2024
Website for White Tree Turkey, an Istanbul-based real estate agency focused on luxury developments and helping clients invest in property across Turkey.
Teen Donut
Offlineteendonut.com · 2022
WordPress site for the Teen Donut coffee-shop & donut chain: menus, locations and brand presence for a growing franchise.
Arezou Mortazavi
Offlinearezoumortazavi.com · 2023
Personal portfolio for a PhD in Cognitive Science & Design: a clean academic showcase of research, writing and design work.
Building structuredAI workflows.
When AI started moving into everyday products, I began learning how to use it in a practical way. I started with workflow automation in n8n, then built RAG systems to connect models with reliable sources and business data. That work gradually led me to agent orchestration.
I now see AI development as bringing specialised agents together, giving each one a clear role, and coordinating them so the system stays efficient, cost aware, and under human control. Through Talero ATS, I also learned to think about security, behavioural audits, traceability, GDPR, and the EU AI Act as part of the architecture, not as checks added at the end.
Multi-agent panel
An eligibility, skills-fit, evidence and uncertainty panel, synthesised by a summariser, runs on pseudonymised profiles.
Deterministic core, AI at the edges
An LLM-free parser stays the source of truth, so ~90% of CV processing costs $0 and every decision is reproducible.
Cheapest-capable routing
Claude Haiku, GPT-4o-mini and Gemini Flash-Lite are chosen per task behind circuit breakers with automatic fallback.
Auditable by design
Every AI decision is logged with engine version, model and an inputs hash, exportable as a regulator-ready PDF.
From websites to AI-native applications.
I began by building websites for small businesses, personal brands, and service companies. Those early projects taught me how to listen to what someone needs, organize unclear ideas, and turn them into something simple and useful.
As the projects became more demanding, I moved beyond WordPress into React, TypeScript, APIs, databases, and cloud deployment. I wanted to understand the full product, not only the part visible in the browser. This led me to work on live platforms, rebuild older systems, and contribute to products used by real users.
My current focus is AI-native software. Through Talero ATS, I have been working on how deterministic software, language models, background processing, and human review can work together in one system. I am continuing to develop in this area by building, testing decisions in practice, and learning from the problems that appear after a system leaves the local environment.
The stack behind the work.
A working toolkit built around AI engineering, with full-stack foundations underneath it.
AI Engineering
01Frontend
02Backend & Data
03Infra & Tooling
04Great products start with a simple hello.
Working on something or just want to trade ideas? Send a note and I'll happily reply.