César Pedro Zea Gomez

I designed and executed the technical enablement and innovation strategy to scale a global data and AI practice from 0 to 25 engineers across four continents. I built the specialized training architecture, internal knowledge platforms, and delivery frameworks required to sustain enterprise-grade production AI.

Key Focus Areas:
• Training and Formation: Designed the end-to-end data and AI training curriculum. Personally supported engineering teams across the full lifecycle, including onboarding, technical formation, and production readiness.
• Centers of Excellence (CoE): Established and managed technical CoEs focused on modern data platforms and MLOps, driving comprehensive certification pathways for Dataiku, MS Fabric, Snowflake, and Databricks.
• Commercial Enablement: Formed a bridge between engineering and business development by training the commercial team on AI positioning and mentoring senior engineers to act as pre-sales technical authorities.
• R&D and Prototyping: Accelerated time-to-market for enterprise clients by standardizing proof-of-concept (POC) workflows and creating scalable architecture blueprints for RAG, LLM applications, and agentic systems.

I built a high-performance data and AI practice from the ground up to align technical execution with enterprise growth strategy. I designed the complete organizational blueprint, secured executive buy-in, and scaled specialized engineering teams from scratch.

Key Focus Areas:
• Strategic Roadmap: Formulated a multi-year vision connecting advanced AI capabilities directly to market differentiation and business value.
• Organizational Architecture: Structured the division from zero, defining roles for specialized squads including Dataiku, MS Fabric, Python AI, front-end AI, and Snowflake.
• Resource Optimization: Authored the hiring strategy to correct misalignments, securing approval for critical hires while restructuring workflows to free senior AI specialists for high-value innovation.
• Talent Acquisition: Directed end-to-end technical recruitment, defining engineering standards and leading the selection process to scale the practice.
• Operational Framework: Established the engineering culture, implementing communication best practices, robust shared documentation, and a framework for execution excellence.

I owned the entire commercial lifecycle and end-to-end delivery execution for the data and AI practice, bridging high-stakes corporate strategy with technical execution. I managed both revenue enablement and technical outcome accountability across a global enterprise client footprint.

Key Focus Areas:
• Commercial Authority: Authored and signed 100% of data and AI Statements of Work (SOWs) and led daily pre-sales operations to defend complex technical architectures before enterprise buyers.
• Enterprise Engagement: Served as the primary technical authority at major industry events, including the Snowflake Summit 2025 and AI Unveiled, closing strategic conversations for Fortune 500 and private equity clients such as Blackstone and Hitachi Energy.
• Delivery Accountability: Maintained absolute ownership of project outcomes across all practice engagements, ensuring engineering delivery aligned precisely with client business goals.
• Hands-on Technical Leadership: Remained close to the keyboard by conducting architecture reviews, writing code, and leading high-stakes troubleshooting to guarantee the execution quality of production AI systems.

I architected and deployed an enterprise-grade retrieval-augmented generation (RAG) and document intelligence platform for banking and financial institutions. I integrated advanced natural language processing into secure financial workflows to enable precise query answering across extensive regulatory and internal document repositories.

Key Focus Areas:
• System Architecture: Engineered a secure, enterprise-ready RAG pipeline utilizing Azure OpenAI and LangChain to process multi-format financial documentation while adhering to strict compliance and privacy standards.
• Document Intelligence: Implemented advanced semantic parsing and context-aware chunking strategies tailored for high-density banking records, minimizing data extraction errors and hallucinations.
• Vector Optimization: Designed vector search and embedding indexing pipelines to guarantee low-latency, highly accurate retrieval under intensive, concurrent enterprise workloads.
• Production Delivery: Guided the complete development lifecycle from initial proof-of-concept to production deployment, ensuring zero-trust security standards and seamless back-end system integration.

As lead architect, I developed an AI-powered platform that transforms raw, unstructured data into actionable business insights using OpenAI's GPT models, giving decision-makers a clear view of customer sentiment, product feedback, and team communications.

Key contributions: a dual-module architecture with specialized modules for product reviews and group chats; a scalable processing pipeline that uses batching to handle text exceeding model token limits; advanced prompt engineering with staging and grounding techniques to minimize hallucinations; a semantic clustering system that identifies dominant topics while keeping references to source text for verification; a comprehensive insights framework generating summaries, key topics, statistical analysis, and recommendations; a natural language query interface enabling non-technical users to explore data; and Azure cloud integration for scalable, secure, compliant deployment.

A comprehensive AI-powered solution that revolutionizes how banking institutions engage customers through personalized product recommendations.

This Next Best Offer (NBO) platform uses machine learning and predictive analytics to identify the most suitable banking products for each customer, calculate expected revenue impact, and generate personalized communications to maximize conversion.

Key contributions: a machine learning pipeline analyzing demographics, account history, transaction patterns, and product holdings to predict subscription probability; multi-faceted customer segmentation combining RFM analysis with behavioral patterns; predictive revenue modeling of direct and indirect ROI per recommendation; an interactive campaign dashboard surfacing top cross-sell prospects; LLM-generated personalized sales communications; a flexible data ingestion system supporting file upload, database connections, or existing segmentation systems; an intuitive web interface for configuring parameters and visualizing results; and an optional 'Advisor' plugin adding deeper customer insights and visualizations for decision-makers.

VoiceRAG demonstrates a powerful pattern for building voice-based assistants with retrieval-augmented generation (RAG), combining modern AI technologies to create natural, context-aware voice interactions enhanced with domain-specific knowledge.

Key contributions: a real-time voice interface using the GPT-4o Realtime API for audio, enabling fluid conversations without turn-based limitations; a sophisticated Real-Time Middle Tier (RTMT) bridging the front end to Azure OpenAI via bidirectional WebSockets; RAG integration with Azure AI Search to retrieve and reference knowledge-base information for more accurate answers; custom financial tools to retrieve product balances, list products, and execute transactions like transfers; a responsive React and TypeScript front end with audio recording and playback, document display, and conversation feedback, styled with TailwindCSS; a secure, personalized architecture authorizing actions by user identity; robust error handling with graceful degradation; and an extensible, scalable design for easily adding new tools. It serves as both a functional prototype for voice-based financial services and a reference implementation of the VoiceRAG pattern.

Solution: Java, Ext JS, and SQL Server Enterprise system software that controls multi-system gateways using a proprietary SOAP protocol and direct pooling, with six specific gateways for each remote control lighting manufacturer. The SQL Server Enterprise database is a 2TB database with partitioned tables containing a billion records.

Performance:
• 3,000+ remote PLCs are near real-time telecontrolled.
• 126,000+ source lights are telecontrolled.
• 4 million+ events and measures are performed each year.
• Near real-time supervision and telecontrol are provided.

Additional accomplishments: I also designed and implemented a BI analytics portal to analyze and verify compliance indicators for the Madrid remote control lightning and ornamental fountains service. As the only developer on this very large and complex project for more than seven years, I delivered all the implementations, maintenance, and evolutions on a part-time basis.

Solution: Copy and keep your live, remote data up-to-date on your server using different SQL Server technologies. I implemented the entire project for Eldon (now nVent HOFFMAN).

SQL Server technologies:
• Ultra-speed data transfers using BCP parallel processes and multi-disks caches.
Three synchronization technologies:
• Amazon AWS-DMS Service is automatically controlled via Amazon API Gateway.
• Partial ultra-speed full update.
• Checksum of each record calculated in each server, transferring only the changes.

Technical details:
• All data processes were programmed in T-SQL.
• Java back end supports the front end and links with Amazon AWS-DMS using the Amazon API Gateway.
• Node.js utilities control quick actions of Amazon AWS-DMS, using the Amazon API Gateway for some purposes not implemented inside the Java back end.
• Ext JS front end.

Solution: A simple and clean solution on SQL Servers without the need to install items that you don't trust. It's perfectly valid for the development or integration of SQL Servers or even, on some occasions, for the control of production servers that are not totally isolated from the developers.

Automatic Git logging of SQL Server T-SQL source code and schema changes happen at the server-side without any action from developers. Al schema changes and all T-SQL source code changes are registered into Git. It is inside the SQL Server where the change is detected and logged, not in the clients. You'll forget that it's installed until the moment you need it.

Key features:
• The changes are detected and recorded by the SQL Server itself.
• All changes are logged without the intervention of developers.
• It's easily adaptable so that the changes made by certain user accounts are not recorded, thus avoiding recording the changes made in the schemas by the BI processes, data warehouse, etc.

Background: The highest load of web stress in Spain, with several million sales per hour, is shortly before each football game. A multilayer infrastructure, front end, and 12 synchronized SQL Servers had been implemented with much of the business layer in hundreds of T-SQL procedures. I carried out the entire project.

Goals: Safe migration, consolidation, and performance improvements in a non-stop production environment. This involved upgrading from SQL Server 2000 to the last SQL Server version, ensuring complete empirical testing and documentation from an engineering perspective.

Requirements:
• 100% verification of the migrated T-SQL code with sufficient unit tests to verify code coverage.
• 100% code coverage, a record of tests performed, and verification of results.
• 100% verification of views and migrated data.
• Increased performance with the same stress and load in production.
• Hot migration performed in production systems.

Results:
• 100% compliance with the project requirements.
• Zero incidents in the hot migration of the systems in production.
• Reduced 12 servers to two servers.
• Completed the project within the budget and other terms with no incident.
• Received congratulatory feedback from the client.