Santiago Vacas

Working on this project has been a very positive and technically interesting experience. During development, multiple backend and frontend challenges were successfully resolved within an architecture based on Python/Django (DRF), React, Docker, and Azure.

Complex integrations with external APIs and Web Services were implemented, including connection with the payroll API for obtaining company cost through a custom authentication flow, as well as the creation of an intermediate layer in Django for bidirectional timesheet synchronization with external time tracking systems.

Additionally, new full-stack functionalities were developed such as an advanced and interactive matrix organizational chart, adapted to non-hierarchical business structures, along with new PDF report generation templates integrated into the existing architecture.

On the frontend, user experience was significantly improved through dynamic tables with advanced sorting optimized for performance and pagination, in addition to refactoring and standardization tasks to leave a cleaner, more maintainable, and scalable codebase.

Communication throughout the project was excellent, allowing for fluid collaboration and rapid resolution of the various technical challenges that arose.

Consulting as a freelance AI Architect on generative AI projects in a corporate environment. Design and implementation of RAG (Retrieval-Augmented Generation) architectures, agentic workflows, and LLM integration into business processes.

Technical work: developing pipelines with LangChain and LangGraph, designing agentic workflows, integrating language models, and optimizing cost/performance in agent and RAG architectures. Developing an application with React (frontend), Python (backend), and PostgreSQL as the database. Collaboration in a corporate environment with multiple technical and business stakeholders.

Stack: Python, React, PostgreSQL, Docker, LangChain, LangGraph, RAG architectures, LLM integration, agentic workflows.

Design and development of an AI conversational voice agent to automate call reception and appointment management for a medical scheduling platform. The agent handles calls autonomously 24/7: it understands the user's request, checks real-time availability against the client's scheduling system (Ofimedic), and confirms the appointment immediately, without receptionist intervention.

Technical work: custom development of an MCP server in Python that connects the agent (ElevenLabs) with the client's existing system via its API (SOAP/REST), managing availability and appointment synchronization, and optimizing the timing between the model's response and audio playback to achieve a natural and fluid conversation. Result: elimination of lost calls during peak hours and complete automation of a process that previously required constant human attention.

Stack: Python, MCP (custom server), ElevenLabs, SOAP/REST APIs, Ofimedic integration.