Lorenzo Germini

•Spearheaded the development of a EdTech web application product, architecting and implementing the backend infrastructure on AWS (DynamoDB, Amplify, Lambda) and Generative AI API layer (OpenAI & Gemini with structured outputs) from the ground up.

•Engineered a robust Retrieval-Augmented Generation (RAG) application utilizing pgai, incorporating custom-built and state-of-the-art parsing pipelines to handle diverse input formats containing both machine-readable and handwritten text.

•Set up LLM tracing (LangSmith) and evaluation pipelines to enable systematic refinement of prompts, LLM outputs and RAG parameters.

•Oversaw the development team in collaboration with the Product Manager, taking on a leadership role in addition to core engineering responsibilities for the MVP, and contributing significantly to product strategy.

•Developed a POC for an anomaly detection system leveraging Sentence Transformers for semantic clustering of maintenance reports and anomaly descriptions.

•Implemented time series forecasting models for predictive maintenance, combining NLP insights with temporal patterns to enhance fault prediction in pharmaceutical manufacturing equipment.

•Explored generative AI use cases, including the use of agents, to optimize production processes and enable intelligent decision-making.

•Led comprehensive analysis of multi-source sensor data for water consumption optimization, implementing Python to identify inefficiencies in Distillators and WFI systems. Designed and managed water rejection strategies projected to save 10,000 m³ annually through combined hardware upgrades and real-time software optimization.

•Created a Streamlit analytics dashboard for clean utilities equipment to monitor KPIs, enabling proactive maintenance and reducing reliance on external companies. This innovation saved costs, cut manual data inspection time by 50%, and earned a silver recognition award for its impact on operational efficiency.

•Developed a proof-of-concept combining gpt-4o-mini and PandasAI to enable natural language interactions with operational data. Built a conversational interface allowing users to query complex datasets through natural language, automatically generating relevant visualizations and data-driven insights.

•Taught Math and Physics at my high-school alma mater, reconnecting with my hometown community. Shared my passion for STEM and fostered a renewed sense of purpose before the next chapter of my career.

•Developed an interior design MVP using Stable Diffusion, implementing Dreambooth fine-tuning on personally curated datasets of room designs. Created custom training datasets through careful selection and preprocessing of interior design examples.

•Led comprehensive market research and user outreach efforts, engaging with interior design firms and potential customers to validate product-market fit. Leveraged insights to make a data-driven decision to conclude the project.

•Developed and deployed a Generative AI tool for biomedical literature analysis, leveraging fine-tuned language models trained on curated scientific datasets. Implemented an end-to-end pipeline for abstractive text summarization.

•Collaborated with the CEO to define product management strategies and streamline business processes for the company's first digital health product, successfully establishing a partnership with a renowned clinic.

•Actively engaged with various business functions to understand user needs and drive impactful solutions, demonstrating a strong commitment to improving health outcomes through technology.

•Managed development of customized filling machine prototype for personalized drug supply, including hardware-software integration testing and collaboration with external partners.

•Improved tuning of filling prototype by creating a Machine Learning model (gradient-boosted tree) to optimize fill parameter determination for Personalized Medicine solutions, reducing tuning time by 20% compared to manual methods.

•Emphasized the practical applications of machine learning to drive innovation in personalized medicine, showcasing ability to translate technical expertise into real-world impact.