Skip to content
Senior AI Engineer

Production AI systems for enterprise and public-sector workflows.

I design and ship agentic systems, enterprise retrieval pipelines, and evaluation harnesses — in Python on Azure, in Arabic and English. Currently Senior AI Engineer at Ghaia.ai.

Based in Islamabad, Pakistan · Open to remote roles and Gulf relocation
Available for selected Senior AI Engineer opportunities
Portrait of Minhal Abdul Sami, Senior AI Engineer

Currently

Shipping multi-agent and retrieval systems at Ghaia.ai for public-sector and enterprise clients in the Gulf.

  • 01Production systems in Arabic and English
  • 02Multi-agent orchestration and enterprise retrieval
  • 03Evaluation harnesses with LLM-as-judge gates
  • 04Enterprise and public-sector deployments
What I can own

Senior ownership across the AI stack.

  • Agentic systems

    Design and ship multi-agent workflows with explicit state, recovery, and evaluation — not freeform prompt chains.

  • Enterprise RAG

    Build retrieval pipelines that survive real corpora — chunking, embeddings, hybrid search, evaluation, and answer grounding.

  • LLM evaluation

    End-to-end suites with LLM-as-judge, regression gates, and human review — trust comes from measurement.

  • Arabic-first AI

    Production Arabic NLP: tokenization, prompt structuring, evaluation, and content generation alongside English.

  • Production delivery

    Python, FastAPI, Azure, Postgres, Playwright. Deploy, observe, debug, and harden under real traffic.

Case studies

Production systems, with the receipts.

Multi-agent content platform

Arabic-first agentic generation for public-sector workflows

Role · AI-layer lead — architecture and implementation

Production multi-agent system generating Arabic-first content for government workflows, with deterministic orchestration and an evaluation suite gating every release.

PythonAutoGenAzure OpenAIWebSocketsPlaywrightPostgreSQL

Award-travel discovery & feasibility engine

LLM discovery + deterministic scoring for a Gulf airline loyalty programme

Role · Designed and evaluated end-to-end

A multi-stage destination-discovery and award-feasibility engine for a Gulf airline's loyalty programme — turning a member's intent into ranked, explainable destination and routing suggestions, benchmarked against a baseline before shipping.

PythonAzure OpenAIHybrid RAGBatched availability APIEvaluation harness

Multi-agent PR intake & routing gatekeeper

Classification, routing, and compliance for public-sector communications intake

Role · AI backend lead — architecture and implementation

Production multi-agent system that ingests inbound communications requests, classifies and validates them against policy, and routes each to the right downstream workflow — behind a strict state machine and data-residency guarantees.

Python 3.12FastAPIMicrosoft Agent FrameworkAzure OpenAIPydantic v2pytest

Arabic generative video pipeline

Multimodal generation with native Arabic audio and burned-in subtitles

Role · POC lead and integration spec author

A generation pipeline that turns a brief into short cinematic video with native Arabic narration and on-screen subtitles — combining video, image, text, and speech models behind one orchestrated workflow.

Azure OpenAIVeo (video)Gemini (image)Azure AI SpeechPythonWebSockets

Enterprise retrieval over Arabic legal corpora

RAG assistant grounded in Saudi legal documents — Al Qanoon

Role · Implemented end-to-end

A retrieval-augmented assistant grounded in Saudi legal documents from the Ministry of Justice, SDAIA, and municipalities — delivered end-to-end as a public-facing assistant.

PythonLangChainAzure OpenAIFAISSMultilingual embeddingsCross-encoder re-ranker
Experience

Shipped roles, in order of recency.

Ghaia.ai

Senior AI Engineer

Oct 2025 — Present

Remote · Doha, Qatar

Architect and ship production multi-agent systems for public-sector and enterprise clients in Arabic and English.

  • ArchitectedArchitected the AI layer of a WebSocket-driven AutoGen system generating Arabic-first government content.
  • ImplementedImplemented a declarative, state-gated prompt-composition engine, replacing a large legacy prompt builder with independently testable sections.
  • OwnedOwned an agentic end-to-end suite with LLM-as-judge scoring for Arabic outputs.
  • ContributedContributed evaluation tooling, retrieval improvements, and production debugging across public-sector and enterprise workflows in the Gulf.

Unisami

AI Engineer

Jun 2023 — Present

Remote

Build custom AI systems for SMB and public-sector clients — RAG, automation, and applied LLM pipelines.

  • ImplementedImplemented Al Qanoon — an Arabic-first RAG assistant grounded in publicly available Saudi legal texts.
  • LedLed delivery of bespoke AI workflows including outreach automation and content generation pipelines.
  • EvaluatedEvaluated retrieval quality and grounding across Arabic and English corpora.

Earlier experience

  • CodingalSenior Coding InstructorJul 2022 — Feb 2025 · Remote
  • Ribat StoreHead of ITSep 2022 — Nov 2023 · Pakistan
  • AdrenodChief Technical OfficerMay 2023 — Aug 2023 · Islamabad, Pakistan
  • WebspireCo-FounderMay 2021 — May 2022 · Remote
About

Pakistan-based Senior AI Engineer focused on production agentic systems, enterprise retrieval, and Arabic-first AI. I work across architecture and implementation — from orchestration and evaluation to backend integration and production debugging.

I prefer systems that are deterministic where it matters and evaluated before they are trusted. Most of my recent work is in production Arabic-first AI for Gulf clients, with retrieval, agent orchestration, and evaluation as the load-bearing pieces.

Capabilities

The stack behind the work.

Agentic systems

  • AutoGen
  • Microsoft Agent Framework
  • LangGraph
  • State machines
  • Tool orchestration
  • Recovery & retries

Retrieval and RAG

  • FAISS
  • Vector search
  • Hybrid retrieval
  • Chunking strategies
  • Re-ranking
  • Citations & grounding

LLM evaluation

  • LLM-as-judge
  • E2E test suites
  • Regression gates
  • Prompt versioning
  • Human-in-the-loop review

Backend & infrastructure

  • Python
  • FastAPI / Flask
  • PostgreSQL
  • SQLAlchemy
  • WebSockets
  • Azure OpenAI
  • Microservices

Applied AI

  • Arabic NLP
  • Document understanding
  • Multimodal generation
  • Workflow automation
  • Public-sector deployments
How I build AI systems
  • Deterministic where it matters

    Wrap LLM calls in deterministic state machines, validators, and retries. The model is one component, not the system.

  • Evaluated before trusted

    No prompt or chain ships without an evaluation suite. Regression gates exist so improvements stay improvements.

  • Retrieval before rhetoric

    Better chunking, hybrid search, and grounding outperform prompt acrobatics in real corpora.

  • Designed for failure

    Timeouts, partial results, graceful degradation. Production traffic is the architecture review.

How I build AI systems

Contact

Let's discuss the system you need to build.

I'm open to Senior AI Engineer roles — remote, or relocation in the Gulf — and to selected build partnerships where the system has to work on Monday, not just demo on Friday.