A department store chain with extensive coverage across Mexico struggled with time-consuming product attribute completion. This process led to incomplete or error-filled product descriptions. The issues negatively impacted the online shopping experience. An AI-powered computer vision system was implemented using Google's Gemini Pro Vision model. It automatically extracted product attributes such as color, material, dimensions, and size from product images. The system also enriched descriptions using historical product data. Text Embedding Gecko was used to retroactively update similar products with newly discovered attributes. The implementation reduced the time required for product attribute input. Product descriptions on the online store were enhanced and became more complete and accurate. Retroactive catalog updates were handled more efficiently. The improved descriptions supported a better customer shopping experience.

Falabella’s manual product description creation was time-consuming and error-prone. Incomplete or inaccurate product information negatively impacted the e-commerce customer experience. The process also delayed platform launches. An AI-powered automation system was deployed to generate and enrich product descriptions. It used Gemini Pro Vision for computer vision analysis and the Text Embedding Gecko model for semantic enrichment. The solution ran on GCP infrastructure with Python and Docker to support scalability. It extracted product attributes such as color, material, dimensions, and size from product images and enriched descriptions with comparative data. The automation processed product templates each day instead of relying on manual creation. It retroactively updated existing products when newly discovered attributes were identified. This approach reduced reliance on error-prone manual steps and supported more complete product information. It also helped address delays tied to description creation during e-commerce launches.

A prestigious private university in Latin America needed to modernize its AI-driven assistant solutions. Previous chatbot initiatives existed, but evolving AI capabilities required an upgrade. The university aimed to optimize response times while maintaining accuracy and scalability. It also sought to improve support for both academic course questions and university policy navigation. Two AI solutions were developed and deployed to address these needs. An Academic Bot was implemented so professors could upload course materials and tailor responses for their classes. A Regulations Bot was implemented to help students and faculty navigate university policies. The solutions were built with FastAPI, Python, Azure, and OpenAI models, using Agile and DevOps methodologies. The deployment enhanced the student experience by providing a reliable, fast AI assistant. It reduced faculty burden by delegating student inquiries to AI. It supported cost optimization compared to hiring additional teaching assistants. Professors were able to personalize bot behavior using course-specific materials.

A major global automotive manufacturer removed hood temperature sensors from one vehicle model to reduce costs. Without sensor data, the team could not proactively detect engine failures occurring in tropical markets. The problem disproportionately affected black-colored vehicles that experienced overheating issues. An Early Fault Detection (EFD) system was implemented using machine learning clustering with Sentence Transformers to analyze customer claim comments. UMAP dimensionality reduction was used to identify fault patterns in the claims data. Vehicle model variables and quality process data were integrated into predictive models. The system was deployed on GCP infrastructure. The approach identified a critical issue tied to engine failures in black vehicles operating in tropical climates. The findings enabled the reintroduction of temperature sensors based on data evidence. Proactive fault detection reduced attrition and claim rate. Real-time analytics supported continuous product innovation.