Emmanuel – Python, LLM, AI agent development

Multi-tenant RAG (Retrieval-Augmented Generation) assistant platform designed for academic and institutional documentation, enabling organizations to manage custom AI bot identities and business rules within a unified infrastructure while delivering accurate, metadata-driven responses using advanced LLMs.

• Designed a dynamic shared-database architecture using Supabase to manage isolated tenant identities and institutional business rules;
• Developed a high-fidelity RAG engine utilizing AWS Bedrock (Claude 3 Haiku & Titan Embeddings) to ensure high retrieval accuracyl
• Optimized semantic search performance by implementing pgvector for efficient indexing of academic documentation;
• Engineered communication workflows with a strategic roadmap for migration to the WhatsApp Business API (Meta) for production-grade scaling;
• Automated code quality and functional validation through GitHub Actions, integrating Ruff for linting and Pytest to certify RAG integrity.

A cloud-native, multi-tenant pharmaceutical and compliance platform designed for the Mexican healthcare market, enabling real-time inventory management and secure handling of sensitive health records across multiple pharmacy branches while ensuring compliance with COFEPRIS (NOM-024) regulations through an event-driven architecture.

• Engineered a fully asynchronous event-driven backend using Python (asyncio) and AWS SQS to decouple core API transactions for horizontal scalability;
• Designed a shared-database multi-tenant architecture in Django with query-level isolation to enable secure cross-branch inventory visibility;
• Developed a custom User Context API with JWT-based stateless authentication to manage branch-level identity and RBAC;
• Implemented a FEFO (First Expired, First Out) algorithm with atomic deduction signals to ensure data consistency across concurrent transactions;
• Architected an immutable, signal-based audit trail compliant with NOM-024 standards, capturing full JSON logs for controlled substance movements;
• Containerized the stack with Docker and orchestrated deployments on AWS ECS, managing IAM roles and environment parity for secure execution.

AI-driven Enterprise Service Management (ESM) ecosystem designed for Fortune 50 clients, automating technical support and incident resolution through LLM orchestration and integration with complex data sources from global MSPs such as Atos and DXC.

• Architected a Historical Incident Ingestion Pipeline: Designed a robust system to ingest over 6 months of historical ITSM data from ServiceNow, Jira, and Zendesk into a Vector Database;
• Engineered 'Agent Assist' (Whisper Mode): Utilized Python and multi-step LLM reasoning to provide real-time, high-fidelity resolution suggestions to support agents, significantly reducing investigation time;
• Re-engineered SharePoint Knowledge Base Ingestion: Leveraged the Microsoft Graph API (Sites.Selected) to enforce strict tenant-scoped access, ensuring security compliance for high-profile enterprise rollouts;
• Led Technical Delivery of Scoped Apps: Managed the end-to-end delivery of multi-domain ServiceNow Scoped Apps, coordinating with external vendors to stabilize global operations;
• Orchestrated Serverless Migration: Shifted operational scripts to an AWS Lambda architecture, achieving zero-downtime deployments and sub-second execution validation with automated CloudWatch monitoring;
• Production Incident Resolution: Spearheaded high-pressure troubleshooting for critical workflow failures during major enterprise rollouts for clients like Microsoft, McDonald's, and Organon.

High-volume crowdfunding and lending platform focused on optimizing the core transactional engine, ensuring financial data integrity, and scaling API infrastructure to support concurrent loan evaluations and complex payment processing workflows.

• High-Performance API Optimization: Refined core REST API endpoints for loan evaluation and transaction lifecycles using Django/DRF, achieving sub-second response times;
• Microservices Context Management: Engineered custom middlewares to streamline user context propagation across a microservices architecture, significantly reducing redundant internal API calls;
• Security Hardening & Identity: Implemented Google OAuth2 and JWT-based authentication with robust Role-Based Access Control (RBAC) to secure sensitive financial records;
• Financial Data Integrity: Managed a high-volume MySQL backend, ensuring strict data consistency across concurrent transaction modules;
• Engineering Excellence: Led rigorous code reviews to enforce PEP8 compliance and Git Flow best practices, improving overall test coverage for mission-critical payment modules.

High-performance image recognition system designed to identify high-risk mosquito-borne disease hotspots, enabling public health authorities to deploy targeted vector control strategies against the Aedes aegypti mosquito.

• Engineered a High-Performance CV Algorithm: Developed a precise image recognition engine using Python and OpenCV to automate the identification of biological vectors for medical research;
• Machine Learning Optimization: Implemented and fine-tuned classification models using Support Vector Machines (SVM) and Naïve Bayes to ensure high diagnostic accuracy;
• Edge Computing Deployment: Optimized the processing engine for Raspbian OS, allowing for real-time analysis on low-power hardware in field environments;
• Strategic Data Generation: Provided the technical foundation to generate heatmaps and high-value data zones, allowing the Health Ministry to identify areas with the highest risk of disease transmission;
• Mathematical Modeling: Utilized NumPy for complex linear algebra and image segmentation to extract critical features from biological samples.