Adrian Simei Tome Lavalle

• Technical Team Training and Development

I led the technical team training in emerging AI technologies, including LangGraph, LangChain, and Google's Agent Development Kit (ADK), strengthening the company's internal capabilities in solutions based on large language models and agentic frameworks. The training program reached 100% of the development team, increasing technical autonomy by 40% and reducing AI project implementation times by approximately 30%.

• Enterprise Systems Implementation and Analysis

I participated in the initial implementation of the Odoo ERP system using LangChain to create intelligent integrations that facilitate the extraction and processing of business information, successfully automating 60% of repetitive system queries. Subsequently, I developed artificial intelligence-powered data analysis initiatives using the OpenAI API, designing solutions that enable automated analysis, natural language processing over enterprise datasets, and generation of actionable insights from structured data, reducing manual analysis time by 70%.

• Document Processing Automation with Deep Learning for Accounting

I developed and trained custom models with PyTorch for automatic classification of accounting documents and structured data extraction. I implemented deep learning architectures specialized in computer vision and document processing that enabled the accounting team to reduce their operational workload by 55% by automating the processing of invoices, receipts, and other financial documents. The system achieved 85% accuracy in data capture, eliminating manual errors and significantly accelerating accounting recording and financial reconciliation processes.

• AI-Powered CRM Solution Design

I collaborated in the proposal and conceptualization of a CRM system powered by artificial intelligence agents built with LangGraph and LangChain on the backend, implementing CopilotKit on the frontend to create interactive conversational interfaces that allow users to interact naturally with AI agents. The architecture utilizes MongoDB as the database for agent state persistence and customer interaction storage. The agents were deployed on the LangSmith platform, which enabled comprehensive testing, real-time monitoring of generated responses, and continuous analysis of agent behavior to ensure system quality and reliability. The solution projects a 45% improvement in customer management efficiency through intelligent automation in processes, including automated tracking, interaction analysis, and personalized recommendations through native React components that communicate in real time with the agents.

• IBM Watsonx Orchestrate for Bnex (Current Project)

I currently provide technical support in the implementation of IBM Watsonx Orchestrate for the client Bnex, working on the orchestration of intelligent workflows for support ticket generation automation. I developed and implemented a complex conversational agent through voice interaction that guides store users through a structured flow for classification and automatic ticket creation according to different problem categories. The agent integrates real-time data validation, user information confirmation, and automated report generation through custom tools, providing a completely hands-free and accessible customer service experience. The implementation has achieved a 65% reduction in average ticket generation time and improved problem classification accuracy by 80%, significantly optimizing the service center's operational efficiency.

• Human Resources Chatbot Development

In parallel, I am developing a chatbot using n8n integrated with LangChain to automate internal human resources inquiries. The system processes requests for information about vacations, permits, and internal policies through conversational agents that understand context and provide accurate responses. The solution has enabled automation of 75% of routine human resources inquiries, reducing average response time from 24 hours to less than 2 minutes and freeing approximately 15 weekly hours of the HR team to focus on higher-value strategic tasks.

Advanced web application serving as an intelligent copilot for sports coaches, optimizing athlete management through data-driven insights. Built with Google ADK using Python the project leverages Google Vertex AI Agent Engines to implement a hybrid AI agent architecture, where a Master Agent orchestrates specialized domain agents (Nutrition, Training, Mental Support) for personalized planning and real-time analysis. Firestore manages comprehensive athlete profiles with seamless synchronization, while secure frontend-agent communication operates through a custom backend implementing OAuth 2.0 authentication and credential management. The integration of Meta WhatsApp Business API enables automated delivery of training routines as PDF documents directly to athletes' mobile devices. The platform delivers automated plan generation, proactive performance analytics, wellness monitoring with injury risk assessment, and integrated clinical referral systems, transforming traditional coaching into strategic, evidence-based athlete development.

• The School of Athens is a philosophical and educational web application inspired by the iconic painting. Developed entirely with Vanilla JavaScript, the project showcases advanced DOM manipulation and dynamic data fetching via the Fetch API. It uses custom web components for a modular and scalable structure, while reactive programming with JavaScript proxies enables automatic UI updates without relying on external frameworks. A simple client-side router ensures smooth navigation, and the integration of Convai's Web SDK provides interactive, AI-driven conversations with historical philosophers.

• The app incorporates Firebase for backend services, such as authentication and real-time database. Complete the UI/UX design process from concept to implementation, including wireframing and prototyping in Figma, demonstrating full-stack design and development capabilities for interactive educational platforms.

• UI/UX Design Mockup in Figma: https://www.figma.com/design/Q7Xk4KAo2mL7NeoZ4ctoyd/Untitled?node-id=0-1&t=Vh6kt0Bmp4PRpJj8-1

• Technical Support & Infrastructure Management: Provided support to over 100 users across multiple regions in Mexico, resolving software, hardware, and network issues.
• Administered and maintained 3 servers, 2 NAS systems, firewalls, workstations, and databases.
• Handled system integration, data migration, and infrastructure management to ensure availability, performance, and security of critical systems.
• Inventory Management Application: Collaborative development with key features such as mass email sending, vacation day inquiries, data export (CSV/Excel), QR code printing, and advanced reporting. Stack: React, Redux, Vite, Bootstrap, and JavaScript.
• Reservations and Payments Platform: Creation of a solution for indoor cycling and weightlifting classes with secure payments via Stripe. Frontend: Vanilla JavaScript; Backend: Node.js and MongoDB. Implemented CI/CD, Docker, and monitoring for high availability.
• UI/UX Design Mockup in Figma: https://www.figma.com/design/wuUAtcmlepRpFNf3pFPQfU/Landing-Page-Wireframe-Desi
• API with TypeScript and Node.js: Development of an API using the repository pattern and PostgreSQL as the database. Focused on layer separation, scalability, long-term maintainability, and integration with various data sources.
• Testing and Code Quality: Implementation of Jest and React Testing Library to ensure high test coverage and quality. Use of linters, formatters (ESLint, Prettier), and continuous integration pipelines for consistency and reliability.
• Automated Fixed Asset Reconciliation: Designed and implemented a solution for automating the reconciliation process of fixed assets between SAP and the fiscal system using Python, pandas for massive data processing, and matplotlib for generating visual reports. Achieved full automation of NOI matching and MOI calculation, reducing processing time from 3 days with a team of 5 people to 2 hours with 2 people, including unit testing to ensure accuracy.