I build production AI systems for international and industrial clients, from LLM agents that automated B2B lead qualification and contributed to 20% revenue growth to edge AI for network gateway diagnostics at Sagemcom Software & Technologies.
At Sagemcom, I deployed anomaly detection and quantized language models on ARM Cortex-A gateways, fine-tuned LLaMA 3 for root-cause analysis, and built a RAG pipeline with Qdrant, Neo4j, and LangChain. I also implemented Docker-based MLOps workflows with CI/CD, MLflow, Weights & Biases, Prometheus, and Grafana.
My work also includes industrial computer vision for quality control, defect detection, surveillance, and real-time sign-language recognition. I take models from dataset preparation and evaluation through API deployment, using Python, PyTorch, Hugging Face, FastAPI, and Docker.
