Skip to content
RR
0

Player 1 ready

Rana Romdhane

I design, train and ship machine learning systems, from predictive models and computer vision to LLM-powered assistants.

Ariana, Tunisia

Drag to spin · Click to burst

▲ Hit a block to reveal a fact

LEVEL 02Character sheet

Meet the player

Engineer by training, researcher by curiosity, builder by habit.

Rana Romdhane

AI/ML Engineer

Ariana, Tunisia

Class
AI/ML Engineer
Guild
ENICarthage, Computer Science
Specialty
Vision-Language Models, RAG, predictive modeling
Side skill
React and TypeScript front-ends

Languages

  • Arabic Native
  • English Advanced (C1)
  • French Upper-Intermediate (B2)

AI/ML Engineer with a Computer Science engineering degree and hands-on experience designing, training, and deploying end-to-end machine learning solutions spanning predictive modeling, computer vision, NLP, and LLM-based applications. Published researcher in Vision-Language Models, combining strong software engineering fundamentals, including modern React/TypeScript front-end development, with a focus on delivering measurable, production-ready results.

High scores

96.81%

Accuracy on compound emotion recognition

BLIP-2 + LoRA

R² 0.997

Forecasting score in production planning

TransportAI

96%

Answer faithfulness of my RAG assistant

TransportAI

87%

Task prioritization accuracy

BERT model

65%

Manual planning time saved

Dabba Dabba

200K+

Financial records analyzed

Maram Company

LEVEL 03Quest log

Experience

Three quests completed so far. Open one to see the objectives I cleared.

  1. TransportAI: an intelligent public transport management platform

    • Designed and delivered TransportAI, an intelligent public transport management platform integrating a React analytical dashboard, an ML prediction engine, an assisted planning system, and a breakdown management module.
    • Benchmarked multiple regression models (Random Forest, XGBoost, Gradient Boosting, Linear Regression) for operational forecasting; selected Random Forest achieving R² = 0.997, enabling data-driven planning from previously underexploited company data.
    • Engineered a multilingual conversational assistant using a RAG pipeline (LangChain, ChromaDB) over a locally deployed LLM via Ollama, achieving 96% answer faithfulness on a 50-question evaluation set.
    • Architected a fully on-premise, cloud-independent stack (React, FastAPI, Oracle 11g, Ollama) and built the platform's responsive React front-end consuming FastAPI REST APIs, guaranteeing data confidentiality and operational autonomy.
    ReactFastAPIScikit-learnXGBoostLangChainChromaDBOllamaOracle
  2. AI-powered task management built on the GTD methodology

    • Developed a BERT-based NLP model reaching 87% accuracy in task prioritization across 10,000+ tasks within an AI-powered task management application (React/Next.js, GTD methodology).
    • Automated scheduling through calendar API integration, cutting manual planning time by 65%.
    BERTHugging Face TransformersPyTorchReactNext.js
  3. Financial data pipelines and executive dashboards

    • Automated ETL pipelines for financial data with Python and Power BI, reducing processing time by 60%.
    • Built 12+ executive KPI dashboards and analyzed 200K+ financial records to surface trends and anomalies.
    PythonPandasPower BISQL

LEVEL 04World map

Projects

Choose a world to explore. Filter by domain or search for a tool.

6 projects shown

LEVEL 05Legendary scrolls

Publications

Research and writing with my name on it.

Conference paperFirst author

2026

Beyond Binary Emotions: A Multi-Architecture Vision-Language Approach for Compound Facial Expression Recognition

COC 2026 international conference (Senegal)

Compares ResNet50, ViT, LLaVA-1.5 and BLIP-2 on the RAF-CE dataset and shows that a LoRA fine-tuned BLIP-2 reaches 96.81% accuracy and 97.91% macro F1 on 14 compound emotion classes.

Magazine articleAuthor

June 2024

Unraveling the Web of Network Protocols

Code Spectrum magazine, IEEE Computer Society Chapter, ENICarthage

A technical article walking readers through how network protocols fit together.

LEVEL 06Training grounds

Education and badges

Where I levelled up, and the badges I picked up along the way.

Sept 2023 – July 2026

Engineering Degree in Computer Science

Diplôme National d'Ingénieur

National Engineering School of Carthage (ENICarthage)

Charguia 2, Tunisia

Sept 2021 – June 2023

Preparatory Studies in Engineering

Physics–Chemistry Track

Preparatory Institute for Engineering Studies of Monastir

Monastir, Tunisia

Badges unlocked

LEVEL 07Inventory

Skills

Pick a bag to see what is inside.

  • Python
  • Java
  • SQL
  • JavaScript
  • TypeScript

LEVEL 08Throne room

Talk to my AI twin

She knows my whole CV and answers in my own voice. Ask anything about my work.

Ready to chat

This is an AI version of me. It answers from my CV, so please confirm anything important with me directly.

Rana's AI: Hi, I'm Rana's AI twin! Ask me about my projects, my research or my experience. You can type, or press the microphone and just talk.

Bonus stage · 3D

Star Dash

Run down the royal road, grab the gold stars and dodge the red bugs. Catch 12 stars in 30 seconds to win 60 coins.

Stars 0/12♥♥♥Time 30s

← → keys, A / D, or tap a side

Bonus stage

Bug Hunt

Every model has bugs before it ships. Squash 8 in 20 seconds to win 40 coins.

Time 20sScore 0Best 0

LEVEL 09Final stage

Let's team up

Have a role, a project or a question? Send a message and it lands straight in my inbox.