Javier – Python, LLM, LangChain

It is a domain-tuned GenAI solution for pharma research and review-heavy workflows. The product helped internal expert teams retrieve, extract, and process knowledge from domain documentation faster and more consistently, reducing manual review effort and improving research operations.

• Designed and delivered retrieval and generation workflows for domain-specific knowledge tasks;
• Improved extraction accuracy and reliability through embedding optimization and context-aware retrieval;
• Built agentic extraction workflows with tool orchestration for multi-step processes;
• Collaborated with stakeholders to turn high-friction manual workflows into production-oriented AI features.

It is a LLM-based systems for healthcare and legal use cases, focused on knowledge retrieval, compliance-oriented workflows, and multi-step execution.

• Built RAG-based systems with memory to improve retrieval quality and iterative task execution;
• Developed compliance-checking workflows using embeddings and domain-aware retrieval patterns;
• Designed agentic flows for multi-step reasoning and task orchestration;
• Contributed to evaluation thinking across retrieval quality and answer quality, rather than treating them as the same problem.

It is an OCR-based document processing solution to extract and structure information from freelancers invoices more reliably. The project also included cloud-based analytics workflows that improved reporting and data-driven decision-making for stakeholders.

• Developed an OCR system that improved data extraction accuracy from 40% to 84%;
• Built ML-based document processing pipelines integrated with cloud analytics workflows;
• Improved data reliability by transforming unstructured inputs into structured outputs for reporting and stakeholder use.

An NLP solution designed to analyze large volumes of customer feedback by identifying recurring topics and sentiment patterns across reviews using clustering methods. The system transforms unstructured customer comments into structured, actionable insights to support decision-making around customer experience and business priorities.

• Вeveloped Python-based NLP workflows to process and analyze customer review data;
• Created topic clustering logic to identify recurring themes in large sets of feedback encoding text via techniques such as TF-IDF;
• Implemented sentiment analysis to detect positive and negative customer patterns using open-source models;
• Helped convert unstructured text into business-facing insights for decision-making and customer experience analysis.