Andrea Polacchini Email & Phone Number

Bologna, Emilia Romagna, Italia

Engaged in the development of machine learning and generative AI solutions.● Developed a generative AI based solution, in particular a RAG (Retrieval Augmented Generation) system relying on ElasticSearch vector database and Watsonx.ai (IBM) language models. It performed a hybrid search on the document base and the most relevant results were summarized with the help of prompt engineering.● Implemented Python backend with FastAPI, following REST convention, calling ElasticSearch and… Show more Engaged in the development of machine learning and generative AI solutions.● Developed a generative AI based solution, in particular a RAG (Retrieval Augmented Generation) system relying on ElasticSearch vector database and Watsonx.ai (IBM) language models. It performed a hybrid search on the document base and the most relevant results were summarized with the help of prompt engineering.● Implemented Python backend with FastAPI, following REST convention, calling ElasticSearch and Watsonx.ai services with the support of LangChain. Containerized the solution with Docker and deployed into IBM Cloud platform in order to make it available to a front end interface for the conversational flow (Watsonx Assistant). Show less

Bologna, Emilia-Romagna, Italy

Engaged in the development and maintenance of an analytical system dedicated to processing sensor data from industrial machines.● Developed a comprehensive ML system for end-to-end processing of time series sensor data from industrial machines, with the specific aim of integrating predictive maintenance features into analytics software used by clients such as Ferrero. Implemented a TensorFlow-based deep learning algorithm for anomaly detection and forecasting. Managed the model lifecycle in… Show more Engaged in the development and maintenance of an analytical system dedicated to processing sensor data from industrial machines.● Developed a comprehensive ML system for end-to-end processing of time series sensor data from industrial machines, with the specific aim of integrating predictive maintenance features into analytics software used by clients such as Ferrero. Implemented a TensorFlow-based deep learning algorithm for anomaly detection and forecasting. Managed the model lifecycle in accordance with MLOps best practices.● The raw data from the machines undergo resampling, preprocessing, and computation of statistical metrics before being visualized within the application.● Managed Python integration and maintenance of a software product, emphasizing adherence to software engineering best practices. Specialized in extending the application by incorporating new machine learning models and associated functionalities. Ensured compliance with REST API conventions within the Django backend framework. Show less

Bologna, Emilia-Romagna, Italy

Expanded a cloud analytics platform to encompass sustainability initiatives.● Evaluated the relevance of Power Automate as a solution for the client's needs and subsequently applied it to automate the robotic process of data workflows, facilitating the computation of sustainability indices.● Conducted a comprehensive exploration and comparison of various data governance solutions, analyzing the pros and cons of each in the context of the current client's requirements.

Berlin, Germany

Designed and implemented a deep learning-based approach for a camera-based traffic monitoring system.● Created a deep learning-based methodology to augment the traditional Kalman filter, specifically tailored for modeling vehicle coordinates derived from a stereo camera system installed in Berlin.● Conducted an in-depth analysis of scientific literature to identify limitations in the classical approach and explored how neural networks could offer utility in addressing these constraints… Show more Designed and implemented a deep learning-based approach for a camera-based traffic monitoring system.● Created a deep learning-based methodology to augment the traditional Kalman filter, specifically tailored for modeling vehicle coordinates derived from a stereo camera system installed in Berlin.● Conducted an in-depth analysis of scientific literature to identify limitations in the classical approach and explored how neural networks could offer utility in addressing these constraints for our specific task. Show less

Bachelor'S Degree, Physics, 110/110 Cum Laude

Master Of Science - Ms, Artificial Intelligence, 110/110 Cum Laude