Personal Portfolio - Achraf Habib

Hi, I'm Achraf, I’m actively seeking an end-of-study internship in AI, Generative AI, Machine Learning, or MLOps, where I can apply my expertise in model deployment, pipeline automation, and real-time AI solutions. .

I studied a Mathematics-Physics- Computer Science pre-engineering program AT INSAT. My hard work and dedication payed off and allowed me to pursue a master's degree at INSAT (National Institue of Applied Science and Technologies).

During this 3-year engineering's program, I built a solid foundation in data science through courses in mathematics, probability, statistical inference, optimization,Cloud and artificial intelligence.

I would describe myself as enthusiastic, driven, curious, and always eager to tackle new challenges🚀.

I'm particularly drawn to the challenge of leveraging LLMs and working on cutting-edge technologies including fine-tuning and Retrieval Augmented Generation (RAG)🌟.

Skills

Below are some of my skills, and always eager to learn more and explore new frontiers 🔍.

Mathematics

I studied a 2-year Mathematics-Physics-Computer Science pre-engineering program. Relevant topics include Linear Algebra, Probability, Statistics, Signal Theory, and Optimization.

Generative AI

I have extensive experience leveraging the powerful capabilities of LLMs and building autonomous agents through Prompt Engineering, Fine-tuning, and Retrieval Augmented Generation (RAG). My expertise includes advanced frameworks like LangChain, LangGraph, TensorRTllm , Ollama , Azure Services.

Artificial Intelligence, Machine Learning

I've worked on different ML/DL algorithms such as CNN and RNN's ( LSTM , GRU) and Advanced Deep learning architectures such as RAFT . I'm familiar with Scikit-learn, NLTK, TensorFlow, Keras, and PyTorch.

Python

Over the past 5 years, I've extensive experience with Python through my engineering studies, academic projects, and professional experience.

Data Engineering

I have experience building real-time data pipelines, leveraging tools such as Kafka, Spark Streaming, Hive, Elasticsearch, and Kibana to ensure efficient data processing, storage, and visualization.

Cloud Platforms

I have experience with Azure. I work with Azure to deploy applications as well as use Azure AI Studio to deploy LLM models. I have a solid foundation in Devops best practise , I'm very familiar with desginig and implementing CI/CD Piplines

Timeline

📌 Below is my professional experience, showcasing my enthusiasm for tackling new challenges!

Financial Market Prediction Models

Part-Time AI Engineer @ Lightray Technologies| Cyprus, Limassol | Remote | Sep 2024 - Present

– Developed financial market prediction models using **LLaMA3.1 405B** to mimic professional investor workflows.
– Established a fine-tuning pipeline with **Snowflake**, reducing training time by 30%, leveraging a multi-GPU environment with **peer-to-peer communication** between GPUs for enhanced efficiency.
– Managed inference with **TensorRT-LLM** and **Triton**, optimizing model deployment.
– Built a custom dataset through web scraping techniques, resulting in a 25% increase in training data volume for new LLMs.

Technical environment: LLaMA3.1 405B, LLaMA3.1 70B, LoRA, Snowflake framework, multi-GPU with p2p communication, TensorRT-LLM, Triton inference server.

Impact: Improved prediction efficiency, enabling more accurate and timely financial insights for the company, and significantly reducing the computational overhead by optimizing model training and inference processes.

Real-time Patient Data Prediction

MLOps Intern @ Applicare.AI| Canada, Montréal | Remote | Jun 2024 - Sep 2024

Fine-tuned **LSTM**, **GRU**, and **Transformer** models on the **MIMIC-III dataset**, improving prediction accuracy by 25%.
Built a web app for real-time patient data prediction, containerized it with **Docker**, and deployed it on **Azure**.
Established an **MLOps pipeline** with **MLflow** for model tracking and versioning, increasing model reproducibility.

Keywords: MLflow, Docker, Streamlit, Azure, GitHub Actions, Azure Container Registry, Azure Container Instance, Azure Monitor, MIMIC Dataset.

The project involved fine-tuning multiple deep learning models to predict patient health outcomes based on historical data, leading to a 25% improvement in accuracy. A web app was developed to provide real-time predictions, making it easier for healthcare providers to access crucial insights.

Impact: The MLOps pipeline streamlined model versioning and deployment, ensuring more efficient management of models and reducing time-to-deployment by 40%. The use of Docker and Azure further enhanced scalability and reliability.

Download Rapport de Stage

AI-Powered Chatbot

Generative AI Intern @ Vermeg | Tunis, Tunisia | Jul 2023 - Aug 2023

Engineered a chatbot to simplify access to Vermeg’s internal **Palmera framework**, using **Langchain’s RAG framework**, improving response accuracy by 25%.
Optimized **document retrieval** and **response time** by 30% using **vector embeddings** and efficient similarity searches with **ChromaDB**.
Architected a user-friendly **Streamlit app** for real-time chatbot interaction, providing an intuitive interface for users to query the system.

Keywords: LangChain, ChromaDB, Large Language Models, Streamlit, RAG, AI Chatbot, HuggingFace.

The chatbot was developed to enhance internal processes at Vermeg, allowing employees to easily access relevant documentation and receive quick responses to their queries. The integration of LangChain’s RAG framework and ChromaDB improved both the accuracy of responses and the efficiency of document retrieval.

Achraf Habib

Skills

Mathematics

Generative AI

Artificial Intelligence, Machine Learning

Python

Data Engineering

Cloud Platforms

Timeline

Financial Market Prediction Models

Real-time Patient Data Prediction

AI-Powered Chatbot

Portfolio

Cold email generator

Deployment of a Containerized Application with Kubernetes on AKS via GitLab CI/CD

RAG ( retrieval augmented generation )

Any Questions? Feel Free to Contact