Enterprise Architecture
Roadmaps, architecture standards, and technology lifecycle decisions, with security considered from the start.
ASHARAF ALI / PORTFOLIODUBAI, UAE
Enterprise AI architecture.
I’m Asharaf. I design agentic systems and private AI platforms for regulated enterprises.
THE ARCHITECTURE ATELIER
Preparing the atelier …
Select a model or use the labels below.
13 YEARS IN ENTERPRISE TECHNOLOGYCONTINUE BELOW
01EXPERTISE
What I work on.
My work spans enterprise architecture and the engineering needed to deploy AI within security, infrastructure, and operational constraints.
AI Domain Architecture
Agentic workflows, model serving, and air-gapped retrieval that fit the available hardware and operating requirements.
Data & Vector Infrastructure
Data platforms and vector stores that scope retrieval to approved corporate data.
Platforms & Frameworks
Backend services in Python and Go, supported by the frameworks and infrastructure needed to run them.
02SELECTED WORK
Systems I’ve
worked on.
A selection of systems, platforms, and decisions across enterprise AI and software engineering.
03APPROACH
Inside the
data boundary.
Where a model runs matters as much as which model you choose. This example follows a request through retrieval, policy checks, private inference, and an audit record.
01 / 04
Retrieval from approved data
Retrieve relevant context from an approved corporate dataset. Keep access scoped to the request and the person making it.
Govern, then build
I establish architecture guidelines, coding standards, and delivery practices early, so teams can use them as the system grows.
Security by design
Session management, permissions, and audit records belong in the architecture. Retrieval is scoped to approved corporate datasets.
Sovereign by default
I design retrieval pipelines and model deployments that operate inside the enterprise boundary, including quantized models on constrained hardware.
Engineering underneath
My architecture decisions draw on implementation work: backend services, data layers, and the constraints of operating them in production.
04FIELD NOTES
From the
working notebook.
Notes on model serving, agent security, and the deployment decisions I’m working through.
All 18 field notesNOTE / 2103 Sept 2026
builderOne drafter, two engines
The same speculative decoder landed in vLLM and llama.cpp in the same week. When two engines that agree on almost nothing both merge it, I read the PRs.
Read the noteNOTE / 2025 Aug 2026
architectYour model weights should outlive the process
Your model weights should outlive the process that serves them. SGLang shipped that idea on August 21.
Read the noteNOTE / 1912 Aug 2026
judgmentWhere do the pixels actually go?
"Multimodal-native" has quietly come to mean a tool call rather than a model property. Qwen's new plugin repo is honest about that; a lot of the commentary around it is not.
Read the note05ABOUT
From software engineering
to enterprise architecture.
I’m based in Dubai. My background is in software engineering, from cloud commerce platforms to microservices and enterprise security systems. I now focus on AI architecture.
That includes agentic workflows, LLMOps, and air-gapped RAG. I define how those systems fit into an organisation, set the engineering standards, and work through the constraints of running models on private infrastructure.
2023
MBA, Technology Business Management
Liverpool John Moores University
2017
BTech, Computer Science & Engineering
University of Kerala
TECHNOLOGIES
Data & Vector
pgvector · ChromaDB · PostgreSQL · Redis · Vector databases
AI Frameworks
LangGraph · vLLM · LangChain · PyTorch · FastAPI
Languages & Web
Python · Golang · TypeScript · JavaScript · React · Next.js
Cloud & Infra
AWS · Docker · GitOps · CI/CD
GET IN TOUCHLinkedIn
Have a project
to discuss?
[email protected]DUBAI, UNITED ARAB EMIRATESBack to the atelier ↑