Jose – Python, LLM, GCP

It's an app designed for doctors to transcribe and store patient visits, with future functionality to support diagnosis through language analysis.

• Built the back-to-end service to do speech-to-text translation using the Whisper model from OpenAI and Cloud Run with WebSockets.
• Integrated the OpenAI API within a chatbot to explain the app's functionality.
• Created a replacement for standard speech-to-text APIs to save costs.

It's a US-based microchip manufacturing company with plants worldwide.

• Integrated Gemini into an internal tool using prompt engineering and implemented parsers to improve search recall, later testing them across various document formats (PDF, PPTX, XLSX, DOCX).
• Helped to train a defect detection algorithm with Paligemma that could detect and segment defects within the images taken in the manufacturing process.
• Created prompts that helped them answer questions more accurately and describe the LLM's thought process.
• Built a fine-tuning pipeline compliant with their privacy restrictions regarding not accessing APIs outside their VPC.

It's a model for one of Mexico's largest retail stores that identifies duplicate clients based on different criteria, such as name, address, email, etc.

• Built a data pipeline with Kubeflow to generate millions of embeddings.
• Created a Vector index for real-time search of similar clients.
• Trained an XGBoost model to find matches between customers.
• Built a REST service served in Kubernetes to do client matching in real time.

It's a document parser for an insurance company.

• Built a custom parser using Google Cloud's Document AI to process specific document types.
• Developed logic to automatically identify and classify documents such as proof of residence and official IDs.

It's a data lake for one of Mexico's largest airlines. This allows management to access data more quickly and allows data scientists to build models faster.

• Created data pipelines that access the database without affecting operations.
• Defined schemas for the BigQuery Data Warehouse.
• Implemented efficient queries for auditing data and building reports.