Vittorio Mussin

Prompt Engineer @Accenture Italia

Milano, IT

Successfully migrated complex prompts across different OpenAI models (from GPT-4o to GPT-4o-mini and from GPT-4o-mini to GPT-4.1-mini), maintaining performance with significantly smaller and more efficient models to achieve a 242% improvement in response latency.• Engineered an end-to-end model evaluation framework, designing entity-based metrics for pre-release CI/CD validation on pull requests and developing an ”LLM-as-a-Judge” pipeline for live monitoring, increasing the total volume of evaluated interactions by 10x while cutting the required manual evaluation effort to three-quarters of its previous volume.• Architected and developed a multi-purpose chatbot for a major telecom company, integrating APIs for real-time user data retrieval and processing.• Developed a production-grade generative AI tool for structured data extraction from PDFs (a key NLP task similar to NER), which included fine-tuning OpenAI models.• RD team researching and implementing novel generative AI for tech adoption.