Skip to content
Available for backend & AI-augmented engineering roles

Yassine Zakhama — Backend architect for the .NET ecosystem, AI-augmented software engineer

Deep roots in C# / .NET — and since 2025, AI at the centre of the workflow: architectural reasoning, code generation, debugging and spec drafting. The fundamentals haven't changed. The speed has.

View my workDownload Resume

Currently at Integration Objects · Tunis, Tunisia

EstimateBot · not a real project

specs/estimate-bot.md

// spec: EstimateBot// in:  "small change, five minutes tops"// out: how long it will actually take// - the answer is never five minutes// - pad by pi, round up, then add a day

A prompt written as a spec. The acceptance criteria are absurd; writing them down before generating is not.

How the loop actually works
Selected work

AI products, and the backends under them

Two MVPs built around Claude and the .NET 10 stack, and two backend systems working the distributed-systems problems that sit underneath anything shipping at scale. Each one says plainly where it stands — one is shipped, the rest are in progress.

Shipped 2026Personal MVP

DeskPilot

AI support triage & reply drafting

Inbound support tickets are classified, entity-extracted and answered with a drafted reply in a single pass. Claude runs behind an ASP.NET Core API with schema-constrained tool output, a per-tenant token budget and a human approval step, so nothing a model wrote reaches a customer unread.

LLM orchestration under real cost, latency and correctness budgets.

  • .NET 10
  • ASP.NET Core
  • Claude API
  • Microsoft.Extensions.AI
  • EF Core 10
  • Redis
  • Wolverine
  • SignalR
Read the case studyShipped
In progressPersonal MVP

Groundline

Grounded document Q&A that refuses to guess

Answers questions over a private corpus and declines anything it cannot cite. Ingestion, chunking and embedding run as background jobs, retrieval is pgvector, and every claim in a response is checked back against a span in a source document before a single token is streamed to the client.

Retrieval quality, grounding enforcement, and evaluation as a test suite.

  • .NET 10
  • ASP.NET Core
  • Claude API
  • PostgreSQL + pgvector
  • Dapper
  • Hangfire
  • SignalR
  • OpenTelemetry
Read the architecture planIn progress
In progressPersonal project

OrderFlow

Event-driven order & fulfillment platform

Order processing built on DDD aggregates and CQRS. Inventory reservation, payment and fulfillment run as asynchronous workflows published through a transactional outbox, with idempotent consumers, retry and dead-letter handling, and order state streamed to clients as it changes.

Aggregate design, transactional consistency and long-running distributed workflows.

  • .NET 10
  • EF Core 10
  • Dapper
  • RabbitMQ
  • MassTransit
  • DDD
  • CQRS
  • SignalR
Read the architecture planIn progress
In progressPersonal project

OpsPulse

Real-time operations & incident platform

Ingests system events at volume, deduplicates alerts into incidents and drives acknowledgement and escalation through an explicit state machine. Writes go through EF Core; dashboards read from Dapper projections and update live over SignalR.

High-throughput ingestion, read/write separation and real-time delivery.

  • .NET 10
  • ASP.NET Core
  • Dapper
  • RabbitMQ
  • SignalR
  • Hangfire
  • Serilog
  • Docker
Read the architecture planIn progress
View all projects on GitHub

My day-job work at Integration Objects is a commercial product under NDA, so it is written up under Experience rather than shown here. Everything on this page is mine to talk about.

The method

The operating layer between a spec and shipped code.

Models write fast. Engineering decides what is worth writing.

Most AI-assisted projects stall because nothing sits between the prompt and production. That layer is the job — literally: AI-assisted development is part of my remit at Integration Objects, and so is validating every generated solution before it reaches the codebase.

Part of the day job, not a side experiment.

What each stage actually does

Before a single line is generated, the intent is written down: inputs, outputs, invariants, edge cases, and the acceptance criteria that will prove it works. A prompt that reads like a spec produces code that can be reviewed like code.

  • Acceptance criteria
  • Invariants
  • Edge cases
  • Constraints

Refine feeds the next spec — the loop closes.

The stack behind the loop — layer by layer

Five layers, each with its own tools and its own failure modes. The top two move fast; the bottom three are where engineering judgement lives.

AI Collaboration

Daily drivers

  • Claude
  • ChatGPT
  • GitHub Copilot

Prompt Engineering

How intent gets encoded

  • Spec-driven prompting
  • Context engineering
  • Chain-of-thought

Code Generation

From spec to scaffold

  • AI-assisted TDD
  • Scaffolding via LLMs
  • Agent workflows

Architecture

The part AI does not decide

  • Clean Architecture
  • DDD
  • CQRS
  • Microservices

Review & QA

Trust, then verify

  • AI-assisted code review
  • Automated test generation

The engineers who will thrive in 2026 are the ones who deeply understand what they're building — because you can't direct an AI toward something you can't define yourself.

Big change is due to the smallest impacts.
Yassine Zakhama
Core engineering stack

The tools I reach for by default

Depth in the .NET ecosystem, with enough breadth either side of it to own a feature end to end.

  • C#C#Primary Language
  • .NET 10Runtime & Base Class Library
  • ASP.NETWeb & API Framework
  • Entity FrameworkORM & Data Access
  • AngularSingle-Page App Framework
  • TypeScriptJavaScript Super Set
  • PythonScripting & Automation
  • JavaJVM Language
  • JavaScriptBrowser & Node Runtime
  • Node.jsServer-Side JavaScript
  • DockerContainerization
  • AzAzure DevOpsCI/CD & Boards
  • GitVersion Control System
  • VSVisual StudioPrimary IDE
  • SQLSQL ServerRelational Database
  • PostgreSQLOpen-source Relational Database
  • MySQLRelational Database
  • ORAOracleEnterprise Database
  • MongoDBNoSQL Document Database
  • InfluxDBTime-Series Database
  • OPCOPC UAIndustrial Interoperability
  • MQTTLightweight Pub/Sub
  • ZMQZeroMQBrokerless Messaging
  • gRPCgRPCContract-First RPC
  • Google CloudCloud Platform
  • ClaudeAI Pair Engineer
Grouped by what it does

Backend

7 in rotation

  • C#
  • .NET 10
  • ASP.NET Core
  • REST APIs
  • Python
  • Java
  • Node.js

Architecture

6 in rotation

  • Clean Architecture
  • Domain-Driven Design
  • CQRS
  • Microservices
  • Clean code & SOLID
  • Design patterns

Data

7 in rotation

  • EF Core / LINQ
  • SQL Server
  • PostgreSQL
  • MySQL
  • Oracle
  • MongoDB
  • InfluxDB

Distributed systems

6 in rotation

  • Event-driven patterns
  • SignalR
  • gRPC / protobuf
  • ZeroMQ
  • AMQP
  • TLS / SSL

Industrial integration

3 in rotation

  • OPC Classic / OPC UA
  • MQTT
  • Azure IoT Hub

Frontend

5 in rotation

  • Angular
  • TypeScript
  • JavaScript
  • HTML / CSS
  • REST API integration

Cloud & delivery

7 in rotation

  • Microsoft Azure
  • Google Cloud Platform
  • Docker
  • Azure DevOps
  • GitHub Actions
  • Git
  • Visual Studio
4+

Years in production .NET

Backend engineering at Integration Objects since Oct 2022

30%

Performance gain delivered

From optimising existing services on the IT-OT platform

6

International distinctions

Awards, supervision and judging across 2014-2020

5

Countries represented

China, Turkey, South Korea, USA and Italy

Experience

Where the systems actually run

Backend engineering since October 2022 on platforms other systems depend on, plus the degree that led into it.

Integration Objects logo

Integration Objects

Current

Software Backend Engineer

Oct 2022 - Present · Tunis, Tunisia

Distributed .NET services and integration components on Microsoft Azure, powering an IT-OT platform that makes disparate IT and operational-technology data sources interoperate securely.

Standing remit
  • Design, build and maintain distributed .NET services and integration components on Microsoft Azure, for scalability, reliability, fault tolerance and efficient service-to-service communication.
  • Design event-driven communication patterns over messaging systems, cutting service dependencies and improving resilience across distributed components.
  • Build and maintain Angular applications and frontend components against backend REST APIs, shipping features end to end across the frontend and the .NET backend.
  • Contribute to architectural decisions and own features and services across their whole lifecycle — design, implementation, testing, deployment, monitoring, troubleshooting and production support.
  • Refactor and modernise existing .NET codebases for maintainability, performance and code quality, applying clean code, SOLID, appropriate design patterns and thorough code review.
  • Bring AI-assisted development into the workflow for code exploration, debugging, documentation, testing and repetitive work — reviewing and validating every generated solution before it lands.
Selected outcomes
  • Raised system efficiency by 30% through new features and products on the platform.
  • Cut database query times by 25% using LINQ to SQL and ADO.NET Entity Framework.
  • Traced memory leaks, CPU spikes and database bugs down to a 40% faster incident response.
  • Streamed real-time data to web dashboards with ASP.NET and SignalR.
  • Moved data between components over ZeroMQ, MQTT, gRPC and protobuf, hardened with TLS/SSL and certificate management.
  • Ran hot, cold and warm redundancy strategies plus a data-recovery mechanism.
  • C#
  • .NET
  • ASP.NET Core
  • Angular
  • Azure
  • Event-driven messaging
  • Microservices
  • SOLID
  • OPC
RedLean logo

RedLean

Web Developer Intern

Jan 2022 - Jun 2022 · Monastir, Tunisia

Designed, built and deployed a Salesforce application spanning a web platform and a mobile app, with AI-powered OCR for automated data detection.

  • Integrated third-party APIs to widen what the Salesforce application could do.
  • Took part in planning and road-map discussions to align delivery with company goals.
  • Salesforce
  • OCR
  • Deep Learning
  • Unit Tests

Anypli

Software Engineer Intern

May 2019 - Jul 2019 · Monastir, Tunisia

Built a web application wrapping Cisco Webex services into flexible, secure virtual meeting and discussion spaces.

  • Contributed 50K+ lines of code to an established codebase via Git.
  • React.js
  • Cisco Webex
  • Docker
  • Webhooks
  • Cloud Integration

Higher Institute of Applied Sciences and Technology of Sousse

Education

Software Engineering Degree

Sep 2019 - Jul 2022 · Sousse, Tunisia

Software engineering, five-year integrated programme.

Higher Institute of Applied Sciences and Technology of Sousse

Education

Preparatory Cycle to Engineering Studies

Sep 2017 - Jun 2019 · Sousse, Tunisia

Mathematics, physics and computer science foundation cycle.

Certificates

  • Microsoft Certified: Azure AI Cloud Developer Associate

    Microsoft · AI-200

  • Introduction to Quantum Computing: Qubit by Qubit

    IBM Quantum

  • Additive Manufacturing — Associate level

    Certification

Languages

  • EnglishFull professional proficiency
  • FrenchFull professional proficiency
  • ArabicNative
Achievements

6 international distinctions, 5 countries

Six years of science fairs, robotics contests and maker camps across five countries — awards won, student projects supervised, and other people's work judged. It is where the habit of explaining a build to strangers came from.

2014–2020
  • Flag of ChinaChina
  • Flag of TurkeyTurkey
  • Flag of South KoreaSouth Korea
  • Flag of USAUSA
  • Flag of ItalyItaly
  1. Flag of China2020

    Teacher & Best Teamwork Award

    Belt and Road Teenager Maker Camp & Teacher Workshop

    Co-hosted by CAST and China's Ministry of Science and Technology, the camp builds cross-cultural exchange, innovation and collaboration between young makers and science educators.

    Chongqing, China · China Association for Science and Technology · Ministry of Science and Technology · November 2020

  2. Flag of Turkey2019

    Project Supervisor · Bronze Medal

    2nd Oğuzhan Özkaya Education Karademir Science Energy Engineering Fair (OKSEF)

    OKSEF is an international project competition; supervised a student project through to a bronze medal.

    Izmir, Turkey · Oğuzhan Özkaya Educational Institution · June 2019

  3. Flag of South Korea2018

    Special Award · Best Project

    MILSET Expo-Sciences ASIA

    Expo-Sciences are non-competitive multicultural exhibitions of STEAM projects by children and youth from around the world.

    Daejeon, South Korea · MILSET · October 2018

  4. Flag of China2018

    International Judge

    China Adolescent Robotics Competition & World Adolescent Robot Contest

    Judged the 18th World Adolescent Robotics Competition — assessing engineering, autonomy and documentation across international teams.

    Guiyang, Guizhou, China · China Association for Science and Technology, with national organiser SEMIA · July 2018

  5. Flag of USA2015

    Honorable Award · 4th Place

    GENIUS Olympiad

    An international project competition on environmental and social issues, hosted at RIT.

    New York, USA · Terra Science and Education, hosted by Rochester Institute of Technology · June 2015

  6. Flag of Italy2014

    Honorable Participation

    Competition of Young Scientists — I Giovani e le Scienze

    The Italian national selection for the EU Contest for Young Scientists, promoting the scientific potential of young people.

    Milan, Italy · Federation of Scientific and Technical Associations, for the European Commission's Directorate-General for Research · May 2014

Full record on the resume
Why work with me

Four reasons this tends to work out

Speed is only useful if what ships holds up. Here is what stays constant regardless of how much of the typing a model does.

Engineering fundamentals first

Design before generation

Domain modelling, boundaries and data flow get decided by a human. Tooling accelerates the build; it does not choose the architecture.

AI as a force-multiplier, not a crutch

Judgement stays in the loop

Every generated line is read, questioned and owned. If I cannot explain it in review, it does not ship.

Production-grade, tested code

Boring where it counts

Tests derived from specs, meaningful logging and handled failure paths — the parts that only matter at 3am.

Clear communication & documentation

Written down, not tribal

Decisions, trade-offs and interfaces documented so the next engineer — or the next model — has the context it needs.

Get in touch

Hiring for a backend that has to hold?

Open to backend and AI-augmented engineering roles, and always happy to talk architecture. I work from Tunis, on UTC+1 all year round.

Direct channels

Email is fastest. Everything else works too.

Or send a message

Three fields. No newsletter, no follow-up sequence.