Skip to content
Sarah Henia.

Available · Tunis · open to relocation

Sarah Henia

AI Solutions & Software Engineer

I build AI systems that run in production, with a human approving every step that matters, and the governance to prove it.

Portrait of Sarah Henia

Read this page as a

Same facts, different emphasis.

AI solutions engineer who turns business needs into governed AI systems that run in production, and can explain them to engineers and executives alike. Designed the governance of three real agentic systems (GDPR, NIST AI RMF, ISO/IEC 42001), managed a product roadmap, and presents in English, French and Arabic.

Solutions & product
Solution architecture, requirements and specifications, roadmap and backlog, Agile/Scrum, stakeholder management, technical documentation, executive presentations
AI governance
AI gateway, guardrails, human-in-the-loop, audit logging, LLM evaluation and metrics, GDPR, NIST AI RMF, ISO/IEC 42001
AI & data
Agentic AI, multi-agent systems, LLMs, RAG, CAG, graph memory, knowledge graphs, MCP, ETL, change-data capture; Azure OpenAI, OpenRouter; SurrealDB, SQL Server, PostgreSQL, Redis, Cosmos DB
Build & cloud
Python, FastAPI, C#, ASP.NET Core, TypeScript, Angular, REST, API gateway, Microsoft Graph API; AWS, Azure, Docker, CI/CD, observability (Logfire, Langfuse)

Systems I built

Three systems, each with a human in the loop

They run inside private company repositories, so the code is not public. The stories are: problem, architecture, decisions, governance and what each one taught me.

  1. CED Tunisia

    CED Tunisia · graduation internship · Feb to Aug 2026

    Collaboris

    A real-time presence library for enterprise web apps, shipped as npm and NuGet packages.

    • Angular 19 library and ASP.NET Core 8 SDK over SignalR and Redis.
    • Presence down to the open dialog, tab or wizard step, across all browser tabs.
    • PresenceAI: a GDPR-by-design journal queried by a governed AI assistant over MCP.

    Read the case study

    1. Browser tabsMutationObserver view
    2. SharedWorkerone socket per browser
    3. SignalR + Redislive presence
    4. CDC journalOLTP to OLAP · Cosmos
    5. Governed AIAzure OpenAI over MCP
  2. The SamurAI

    The SamurAI · for a Middle East cybersecurity client · 2026

    Market-intelligence system

    A multi-agent system that reads the market in English and Arabic and proposes leads to a human.

    • Allow-listed sources, bilingual ingestion, schema-enforced LLM extraction.
    • One knowledge graph: every claim traces back to its source document.
    • One AI gateway for every model call; human review before any CRM write.

    Read the case study

    1. Sourcesallow-listed, EN / AR
    2. Ingestfetch · clean · store
    3. Classify + extractschema-enforced LLM
    4. Knowledge graphone company, one node
    5. Human reviewbefore any CRM write
  3. The SamurAI

    The SamurAI · in production since March 2026

    Dawn

    An autonomous project-management agent. Nothing is written until a human approves it.

    • Turns recorded meetings into tasks with owner, priority and due date.
    • Graph memory in SurrealDB with hybrid RAG and a Redis context cache.
    • Human approval by email before any write to Microsoft Planner.

    Read the case study

    1. Meetingsrecorded transcripts
    2. Graph memoryRAG + Redis CAG
    3. LLM extractionowner · priority · due
    4. Human approvalevery task
    5. Microsoft Plannervia Graph API

Stack

What I build with

The same wall as on my GitHub profile. Each system page lists its own.

AI

  • Claude
  • OpenAI
  • MCP
  • LangChain
  • Hugging Face
  • Pydantic

Build

  • Python
  • TypeScript
  • C#
  • .NET
  • FastAPI
  • Node.js
  • GraphQL

Front

  • Angular
  • React
  • Next.js

Data

  • SurrealDB
  • PostgreSQL
  • SQL Server
  • MongoDB
  • Redis
  • Neo4j

Cloud

  • AWS
  • Azure
  • Azure DevOps
  • Docker
  • GitHub Actions
  • Nginx
  • Linux

Journey

From business intelligence to shipping governed AI

  1. ESSECT, Tunis · 2020 to 2023

    Bachelor's degree, Business Intelligence

    Where I learned to read data before writing code: modeling, SQL, reporting, and how a business actually asks questions.

  2. ESPRIT, Tunis · 2023 to 2026

    Engineering degree, Software Engineering

    Full-stack JavaScript first, then .NET, Python and cloud. The academic projects below come from these three years.

  3. 2024 to 2025

    Cisco CCNA, AWS Cloud Practitioner, NVIDIA DLI

    Cisco CCNA for networks, AWS Cloud Practitioner for the cloud, and four NVIDIA Deep Learning Institute courses: Adversarial Machine Learning, Building RAG Agents with LLMs, Fundamentals of Deep Learning, and Applications of AI for Predictive Maintenance.

  4. The SamurAI · 2025

    Product Manager, Dojo

    Managed the roadmap and backlog of Dojo, an AI-powered platform that generates on-demand test labs, turning stakeholder needs into specifications and sprint priorities.

  5. CED Tunisia · Feb to Aug 2026

    Software Engineering Intern, Collaboris

    Built Collaboris end to end in three releases under department-head review, and presented the architecture, governance and business case to CED executives.

  6. The SamurAI · 2026 to present

    Software Architect & AI Solution Integrator

    Designing the market-intelligence system and its governance: one graph, one gateway, one human.

Earlier work

Academic projects, 2023 to 2025

Built at ESPRIT, usually in teams and in sprints. Smaller in scope, but this is where the habits came from.

About

Software engineer. AI solution integrator.

I design and build systems where a model proposes, a person decides and a database remembers why. Right now that is Dawn, in production since March 2026, and the market-intelligence system in pilot. Before that it was Collaboris, delivered in three releases at CED Tunisia.

I came to software through data: a business-intelligence degree first, then software engineering at ESPRIT. It shows in how I work. I care about what a number means as much as about the code that produces it, and I write the architecture document before the first commit.

I work across the stack: Python and FastAPI, C# and .NET, Angular and TypeScript, SurrealDB, Redis and SQL Server, deployed on AWS and Azure. I present in Arabic, English and French, to engineers and to executives.

  • Based in

    Tunis, Tunisia · open to relocation

  • Languages

    Arabic (native) · English (C1) · French (B2)

  • Education

    Engineering degree in Software Engineering, ESPRIT (2026) · Bachelor's in Business Intelligence, ESSECT (2023)

  • Certifications

    Cisco CCNA (2024) · AWS Certified Cloud Practitioner (2025)

  • NVIDIA Deep Learning Institute (2025)

    • Adversarial Machine Learning
    • Building RAG Agents with LLMs
    • Fundamentals of Deep Learning
    • Applications of AI for Predictive Maintenance

Contact

Let's talk about what you are building

Solutions engineering, AI governance, AI engineering or software engineering roles. Europe, the Middle East and the Americas.