AI/ML Engineer & Software Engineering Student building intelligent systems that bridge research and production.
I'm a 3rd-year Software Engineering student at the National Institute of Applied Sciences and Technology (INSAT) in Tunis, Tunisia. My passion lives at the intersection of AI research and real-world deployment.
Currently interning at VistaDeep, I work on end-to-end ML systems — from data pipelines and model training to containerized microservices in production. I specialize in LLM engineering, RAG architectures, and quantitative ML.
When I'm not training models, you'll find me contributing to the Google Developer Group INSAT community or exploring the latest advances in AI research.
Software Engineering — INSAT
2023 – PresentAI/ML Intern — VistaDeep
CurrentArabic · French · English
GDG INSAT · Interact Club
Working on Artificial Intelligence & Machine Learning projects including RAG systems, quantitative trading platforms, NLP models, and anomaly detection systems. Building end-to-end ML pipelines from research to production deployment.
Led media operations and content strategy for the club, managing communications and community engagement initiatives.
An AI-driven quantitative trading platform combining XGBoost forecasting and PPO-based portfolio optimization. Trained on 15 years of financial data (56k+ samples, 94 features) achieving up to 80% directional accuracy and a Sharpe ratio of 5.47.
class FinXPredictor:
def __init__(self):
self.model = XGBRegressor()
self.agent = PPO("MlpPolicy")
self.features = 94
async def predict(self, data):
forecast = self.model.predict(data)
action = self.agent.predict(state)
return {
"direction": forecast,
"allocation": action,
"latency": "87ms"
}End-to-end Retrieval-Augmented Generation architecture with document embeddings, FAISS vector search, and LLM response synthesis. Deployed as a FastAPI microservice with prompt engineering and context filtering to reduce hallucination.
Fine-tuned DistilGPT-2 (82M parameters) on 10,000+ arXiv papers achieving 31.5 perplexity. Features interactive Gradio interface and mixed-precision training for optimization.
Unsupervised anomaly detection system trained on 1M+ HDFS log sequences using autoencoders. Full preprocessing and PyTorch pipeline with reconstruction-error thresholds for high-precision anomaly detection.
Comprehensive EDA on 50 years of environmental data (148+ indicators) including CO₂, CH₄, N₂O, and PM2.5. Trend analysis and spatial mapping delivering reproducible workflows and policy-relevant insights.
I'm currently looking for new opportunities in AI/ML engineering. Whether you have a question, a project idea, or just want to say hi — my inbox is always open.
Say Hello