Welcome to my GitHub profile!
I’m a Junior AI/ML Engineer, recently graduated from the National Institute of Applied Sciences and Technology (INSAT) (top 3 of my class), specializing in Networks and Telecommunications with minors in Machine Learning, Deep Learning, Data Science, Computer Vision, and NLP.
I have 2+ years of professional experience in AI, Generative AI, and LLMs, building solutions for industrial scene understanding, intelligent chatbots, conversational assistants, and advanced speech-to-speech interactive agents. My passion lies in advancing open-source LLMs, fine-tuning models for efficiency, and applying AI to real-world challenges in industry.
- ⭐ Ranked in the top 3 of my academic class.
- ⭐ Designed and deployed an AI agentic workflow for 3D industrial scene understanding at SAMP (Paris).
- ⭐ Built LLM-powered agents, RAG pipelines, and Generative AI chatbots at iTransformer365.
- ⭐ Developed an advanced speech-to-speech multimodal assistant by combining ASR, TTS, and LLM-driven dialogue.
- ⭐ Led bilingual transcription projects (WhisperX, Nemo, Pyannote) at Innovative Realities (England).
- ⭐ Built and deployed sentiment analysis systems at ESG Smarter (Canada/Tunis).
- ⭐ Benchmarked advanced models for the ASVspoof Deepfake Detection competition.
- Languages: Python, Matlab, JavaScript, TypeScript, C, Java, SQL, Bash
- Frameworks & Libraries: PyTorch, TensorFlow, Hugging Face, LangChain, LangGraph, OpenCV, FastAPI, Node.js, Angular
- ML Techniques: Regression, Classification, Clustering, Feature Engineering, Model Evaluation, Hyperparameter Tuning
- DL Architectures: CNNs, RNNs, GRUs, LSTMs, Transformers, Attention Mechanisms, Vision Transformers (ViT), YOLO, ResNet
- LLM Fine-Tuning: LoRA, QLoRA, PEFT, instruction tuning, domain adaptation
- Open-Source Models: Mistral, Qwen, LLaMA, DeBERTa, BERT, T5, WhisperX, FasterWhisper, Nemo, Pyannote
- Deployment: Worked with RPoD (RunPod) and Ollama server for efficient local and cloud model hosting
- RAG Pipelines: Retrieval-Augmented Generation for knowledge-grounded responses
- Prompt Engineering: Advanced structured prompting, multi-step reasoning
- Agentic Workflows: Modular multi-agent systems with LangChain, LangGraph, MultiChain
- Intelligent Chatbots & Virtual Assistants: LLM-driven assistants with context-awareness
- Speech-to-Speech Agents: ASR + TTS + LLM dialogue for natural multimodal conversations
- Conversational Intelligence: Dialogue management, memory and context management, emotion-aware responses
- Generative Conversational AI: Task-oriented and open-domain conversational systems
- Memory Handling: Long-term memory, vector-store backed memory, and conversation state tracking for continuity
- 3D AI & Spatial Understanding: 3D Point Clouds, Digital Twins, Spatial Grounding with LLMs
- Audio & Speech Processing: Diarization, speech enhancement, bilingual transcription
- Data Science: Preprocessing, Cleaning, Augmentation, Feature Extraction, Visualization
- DevOps Tools: Git, Docker, Docker Compose, CI/CD, GitHub, GitLab
- Cloud Platforms: AWS & OVH (H100 GPUs, scalable training/deployment), Azure ML, Azure Virtual Machines
- MLOps: MLflow, Weights & Biases (W&B) for experiment tracking, deployment monitoring
- Deep Learning Specialization — DeepLearning.AI
- Machine Learning Specialization — DeepLearning.AI
- Agent in LangGraph — DeepLearning.AI
- Effective MLOps — Weights & Biases
- AZ-900: AI Fundamentals — Microsoft
- AZ-900: Azure Fundamentals — Microsoft
- CCNA: Introduction to Networks & Switching, Routing, and Wireless Essentials — Cisco
- Fortinet NSE 1, 2, 3 — Fortinet Network Security Expert
- JavaScript Algorithms and Data Structures — freeCodeCamp
- Email: [email protected]
- LinkedIn: Haykel Bargougui
- GitHub: HaykelBargouguy