AI Engineer · Software Engineer · Sfax, Tunisia

Mohamed Amin MAKNI

AI Engineer building systems that have to be right in production, not only in evaluation

A model that scores well and a system a team can rely on are different things. I build the second, and I test the parts that decide whether it can be trusted.

Explore my work Contact me
LinkedIn GitHub Sfax, Tunisia
Open to full-time rolesAI Engineer · Applied AI / Machine Learning Engineer — Based in Sfax · Open to Tunis, remote, or relocation. Available immediately.
01Applied AIVision · OCR · Evaluation
02Generative AIRAG · LLMs · Agents
03AI product deliveryBackend · APIs · MLOps
MODEL · DATA · FEEDBACK · DELIVERY
Portrait of Mohamed Amin MAKNI
FocusApplied AI
& intelligent software

01 / Positioning

Engineering useful AI across models, software, and human workflows.

AI Engineer working across computer vision, document intelligence, RAG and agent systems. I work the full path — dataset preparation, model training and evaluation, backend and API integration, business rules, and the human review paths that catch what the model gets wrong. I also ship complete products: Fytrak is a three-role coaching platform I built end to end, alone.

Secure by designObservable in operationMeasured by outcomes

02 / Experience

Work shaped by real workflows

Selected engagements that demonstrate system ownership, backend delivery, and applied engineering.

01

Feb – Jul 2026

Sfax, Tunisia

Essilor SIVO (EssilorLuxottica)

AI & Software Engineer — Final-Year Engineering Project

Designed and delivered OptiFlow Precal Insight, a mobile and AI workflow for processing PRECAL optical forms, now deployed at SIVO.

  • Owned requirements analysis, software and AI architecture, mobile and backend implementation, and end-to-end integration.
  • Built a multi-model document pipeline with confidence scoring, business validation, and human correction loops.
  • Validated the system through model evaluation and end-to-end human-in-the-loop workflow checks.
YOLO11EfficientNetV2-STrOCRGLM-OCRRapidFuzz
02

Jul – Aug 2025

Remote · Paris, France

Altavo Partners

Backend Developer Intern

Built the bank and arbiter side of Hex-Port, a Hedera trade-finance platform for African exporters.

  • The bank service and its REST controllers: KYC review, document validation, dispute handling, bank authentication and order approvals.
  • The approvals that release escrowed payment in two tranches, on shipment and on delivery.
  • Every order state change is written to a Hedera consensus topic, so the audit trail is verifiable rather than asserted.
TypeScriptNode.jsExpressPrismaPostgreSQLHedera SDKSolidity
03

Jul – Aug 2024

Sfax, Tunisia

Infotech Consulting Services (ICS)

Full-Stack Developer Intern

Built REST endpoints for the examinations and teaching-assignment side of a university administration platform.

  • Exam sessions, invigilation, teacher workload and timetables, student enrolment.
  • Wrote the API documentation the rest of the team integrated against.
LaravelVue.jsPHPMySQL

03 / Expertise

A focused engineering toolkit

Capabilities are grouped by the systems they help deliver—not presented as an undifferentiated technology wall.

01

Applied AI Engineering

Model pipelines designed around measurable application outcomes

  • Python
  • PyTorch
  • TensorFlow
  • YOLO
  • OCR
  • Model evaluation
02

Generative AI Systems

Grounded AI workflows for retrieval, reasoning, and assistance

  • RAG
  • LLMs
  • Intelligent agents
  • Vector search
  • Human-in-the-loop
03

Computer Vision & Documents

Detection, recognition, classification, and validation workflows

  • YOLO11
  • TrOCR
  • GLM-OCR
  • EfficientNetV2
  • RapidFuzz
04

Backend & Architecture

Maintainable APIs, services, data flows, and integration boundaries

  • FastAPI
  • Node.js
  • REST APIs
  • PostgreSQL
  • MySQL
  • MongoDB
05

AI Software Delivery

Containerized AI services and repeatable engineering workflows

  • Docker
  • Linux
  • Git
  • GitHub
  • API integration
06

Full-Stack Engineering

Product interfaces connected to production-oriented services

  • Next.js
  • React
  • Angular
  • TypeScript
  • WebSocket

04 / Selected work

Proof through engineering decisions

Selected for relevance, architectural depth, and evidence I can show.

Personal project · Shipped solo · Mobile platform

Fytrak

A three-role coaching platform — trainee, coach, admin — built end to end on my own: mobile app, admin console, cloud backend.

Role
Sole engineer, end to end
Challenge
Give a coach and their client one place to run a programme together, with an access model strict enough that neither can see what the other should not.
3Roles
3Languages
TestedSecurity rules
TypeScriptReact Native (Expo)FirebaseCloud FunctionsFirestorei18n / RTL
Source on GitHub
  • Fytrak trainee home screen showing a pending coach request and the day's calorie and protein targetsTrainee home — targets, and a coach request still pending
  • Fytrak coach directory listing coaches with specialisms, ratings and client countsTrainees browse coaches and send a request
  • Fytrak workout logging screen with a personal-record notification derived from the logged setsA PR is derived from the logged sets, not entered by hand
  • Fytrak meal search showing Tunisian dishes with Arabic names and calorie valuesMeal logging on a Tunisian food database
  • Fytrak coach view confirming that a prescribed workout has been assigned to a client's dayCoach side — a prescription lands in the client's day
  • Fytrak coach reading a client's daily report with nutrition summary and completed sessionand the coach reads back what actually happened

Confidential case study · Document intelligence

OptiFlow Precal Insight

A mobile and AI system that turns photographed PRECAL optical forms into reviewed, business-ready data.

Role
End-to-end software and AI engineering
Challenge
Replace a slow, error-prone manual transcription flow while preserving expert control over uncertain predictions.
92.66%Precision
92.79%Recall
95.16%mAP@50
YOLO11EfficientNetV2-STrOCRGLM-OCRRapidFuzzHuman-in-the-loop
View case study

Altavo Partners · Team of five · Blockchain trade finance

Hex-Port

A trade-finance platform for African exporters where a bank arbitrates a three-party escrow on Hedera, and every state change is written to an immutable audit trail.

Role
Bank and arbiter side, end to end
Challenge
A buyer will not pay before shipment and a seller will not ship before payment. Put a bank in the middle, and the software has to be trustworthy enough that neither side has to trust the other.
3Parties
2 tranchesRelease
HCSAudit trail
TypeScriptNode.js / ExpressPrismaPostgreSQLHedera SDKSolidity

Public repository · RAG & computer vision

SmartWarehouse AI

A logistics platform that combines vehicle-image processing, grounded AI decisions, and warehouse operations.

Role
Full-stack and AI engineering
Challenge
Connect inbound and outbound warehouse workflows to vision, inventory data, and context-aware assistance.
FastAPINext.jsYOLOGeminiChromaDBMySQL
View repository

Team of 4 · In progress · Agent infrastructure

ORYS — AI layer

The AI layer of ORYS, a multi-tenant business-management SaaS for small companies — invoicing, stock, clients, staff. I am responsible for this layer; the platform is built by a team of four.

Role
Responsible for the AI layer
Challenge
Let an assistant take real actions inside a business system without ever becoming a privileged database user, and without depending on a single model vendor.
TypeScriptAnthropic SDKOpenAI SDKZodVitestNestJS

Multimodal AI · Learning analytics

Intelligent Online Classroom

A multimodal learning system connecting quiz responses, webcam data, facial analysis, and NLP-driven interaction.

Role
Full-stack and AI engineering
Challenge
Synchronize behavioral and textual learning signals while keeping the application workflow secure and usable.
AngularPythonYOLOTensorFlowNLPComputer Vision

05 / Engineering approach

From problem framing to operational feedback

01

Frame the problem

Clarify the workflow, constraints, users, and definition of success.

02

Design the system

Define service boundaries, data paths, security controls, and failure modes.

03

Build for delivery

Implement maintainable services and automate repeatable build and release work.

04

Observe and improve

Validate outcomes, expose useful signals, and iterate from operational feedback.

06 / Credentials

Education and verified certifications

Education

Software Engineering & Business Intelligence

National Engineering Degree in Computer Engineering

International Institute of Technology (IIT) · Sfax, Tunisia

Completed · 2021–2023

Preparatory Classes for Engineering Schools

Preparatory Institute for Engineering Studies of Sfax (IPEIS) · Sfax, Tunisia

Certifications

14 October 2024

Information Technology Specialist — Artificial Intelligence

Certiport · Score 889 / 1000

June 2025

Hashgraph Developer

The Hashgraph Association

27 March 2024

Information Technology Specialist — Python

Certiport · Score 880 / 1000

07 / Engineering notes

How I think about building these systems

Short write-ups on evaluating document-AI pipelines, designing review steps people can use, and what I would do differently next time.

Read all notes

08 / Contact

Let’s build systems that are ready to operate.

For AI engineering, computer vision, RAG, intelligent-agent, backend, or software architecture opportunities, reach me directly.

Open to full-time roles: AI Engineer, Applied AI / Machine Learning Engineer. Based in Sfax · Open to Tunis, remote, or relocation. Considering Tunisia · European Union · United Kingdom · UAE · Remote.

mki.medamin@gmail.com
LinkedInGitHubSfax, Tunisia
Download CV — EnglishTélécharger le CV — Français