Manuel Romero's Work | ContraWork by Manuel RomeroManuel Romero
Agentes de IA: Devuelvo el tiempo y la calma a tu negocio
Enterprise AI: Secure MS Teams Ticket Management Bot
Developed a secure AI agent for Microsoft Teams to automate incident management and technical support through intelligent ticketing.
Achievements:
• Security: Integrated OAuth 2.0 and RBAC for secure access within the corporate tenant.
• API Orchestration: Connected to internal APIs for real-time incident retrieval and ticket injection.
• Bot Framework: Built with Azure Bot Framework for multi-turn, context-aware dialogues.
• Automation: Designed a classification flow to route issues to the correct department via API.
Improved support efficiency via a 24/7 self-service tool for incident reporting.
Due to NDA and data privacy policies, proprietary code and internal UIs are not disclosable. The attached architecture diagram illustrates the technical framework implemented. Automated Document Intelligence for B2B Loyalty
Description:
I engineered an end-to-end pipeline to automate data extraction from 500+ heterogeneous invoices/month, directly fueling a corporate loyalty program.
Technical Achievements:
• Extraction: Deployed Azure Document Intelligence with custom models for complex layouts.
• Architecture: Built a serverless backend with Azure Functions for validation and DB injection.
• Integrity: Implemented Fuzzy Matching and SKU cross-referencing for 99% accuracy.
• Semantic Matching: Developed logic to identify promo products within dense invoice data.
Result: Eliminated manual entry, reducing processing time by 85%.
Due to NDA and data privacy policies, proprietary code and internal UIs are not disclosable. The attached architecture diagram illustrates the technical framework implemented. Enterprise RAG: Data Fusion at Scale
I engineered a high-performance RAG pipeline to unify 3 databases (18,000+ records) and 200+ pages of catalogs into a single conversational engine.
Technical Achievements:
• Data Fusion: Used GPT-4o to format and merge disparate sources into a unified, AI-ready schema.
• Scalable Indexing: Orchestrated Azure Functions to automate the merge-index process into Azure AI Search.
• Hybrid Retrieval: Implemented semantic search with high-dimensional embeddings for surgical accuracy in product discovery.
• Optimized Stack: Deployed via Azure Static Web Apps and API Apps, using GPT-4o-mini for fast, cost-effective responses.
Result: A scalable, enterprise-grade assistant that understands complex technical specs and real-time inventory with minimal latency. n8n Local Assistant Demo v0.1.1 - Consultant Agent (In progress)