Agustina Rocca Silvero

🔹 Project: AlexandrIA Platform – GenAI Assistant Development

Design and development of intelligent assistants based on LLMs using RAG architectures and semantic prompts linked to ontologies. Implementation of GenAI solutions in educational and training contexts. Involved in functional validation, performance testing, and technical deployment of assistants.

🔹 Project: Responsible AI Governance – Audit and Regulatory Compliance

Technical audit of AI systems aligned with GDPR, the EU AI Act, and ISO/IEC standards. Implementation of risk controls such as traceability, human validation, access management, and bias detection. Development of adversarial testing and red teaming on generative models. Drafting of compliance documentation: incident response plans, governance frameworks, and impact assessments.

Technologies: Python, Azure OpenAI, AWS, LangChain, FAISS, Neo4j, OWL, LLMs, Prompt Engineering, RAG, RAI, ISO/IEC 42001

🔷 Project: AI Applied to Marketing and Product – Natural Language Processing

Development of AI solutions applied to market analysis, image classification, automatic generation of product descriptions, and personalization systems. Design of multilingual sentiment analysis models.

Development of user profiling systems based on automated web data extraction and behavioral analysis using clustering techniques. Implementation of RAG architectures to extract dynamic insights from product databases and customer reviews. Design of intelligent chatbots for personalized marketing campaigns and user support.

Technologies: Python, Transformers (Hugging Face), RAG, NLP, Web Scraping, Image Classification, Chatbots

🔷 Project 1 – Book Recommendation System Using Collaborative Filtering

Development of a recommendation system based on collaborative filtering with cosine similarity, applied to previous user ratings of books. Construction of a user–item matrix and calculation of user similarities to identify affinities and generate personalized recommendations.

🔷 Project 2 – Automatic Generation of Real Estate Descriptions Using ChatGPT-3

Prompt engineering applied to ChatGPT-3 to automatically generate real estate property descriptions. Prompts were specifically designed to match the unique characteristics of each property (number of rooms, location, condition, etc.), generating relevant and engaging texts for publication.

Technologies: Python, Cosine Similarity, Collaborative Filtering, Recommender Systems, ChatGPT-3, Prompt Engineering