Desarrollador de software - Airtrace Technologies SL - Madrid, España
(2023-10-01 - 2025-10-31)
Data Scientist and Machine Learning Engineer with experience in automation, cloud deployment (AWS), and RAG system development. Skilled in building end-to-end ML pipelines, optimizing model performance, and integrating AI systems into production workflows.
- Performed comprehensive data analysis, including distribution studies, filtering, correlation assessments, and normalization techniques to prepare clean, high-quality datasets for ML models.
- Trained, fine-tuned, and optimized models such as XGBoost, GDN, and STGNN, improving predictive accuracy and efficiency.
- Utilized Weights & Biases for large-scale hyperparameter tuning, monitoring performance metrics to identify the best configurations across multiple experiments.
- Designed and implemented automated workflows using n8n, integrating APIs, data sources, and AI models to streamline operations.
- Developed Retrieval-Augmented Generation (RAG) systems, managing data retrieval, embedding, and contextual augmentation for enhanced generative AI performance.
- Containerized applications with Docker for consistent, portable deployments, optimizing images by including only essential libraries.
- Deployed and managed solutions on AWS, including S3, EC2, SageMaker, API Gateway, and Lambda, ensuring scalable, cost-efficient, and secure environments.
- Built and exposed RESTful endpoints for real-time inference and service integration.
- Validated and tested APIs using Postman, ensuring reliability of GET/POST requests and seamless data ingestion.
- Integrated and validated IoT data streams via ThingsBoard, enabling dynamic data visualization and device monitoring.
- Created and maintained GitHub repositories, managing commits, branches, and code reviews to ensure code integrity and collaborative efficiency.
- Authored detailed technical documentation covering project workflows, data pipelines, and model deployment procedures.
- Delivered client presentations and live demonstrations, effectively translating complex technical results into actionable business insights.
