Sergio García

Design and development of production-grade AI systems leveraging Large Language Models (LLMs) to improve knowledge access and productivity in large-scale enterprise environments, supporting thousands of internal users.

• Designed and implemented RAG and agent-based systems integrating enterprise tools and data sources (via MCP and APIs)
• Built semantic retrieval pipelines using vector-based search for large-scale documentation and codebases
• Developed multi-agent architectures for task orchestration, coordination, and decomposition
• Addressed LLM context limitations by designing scalable strategies using sub-agents and context compaction
• Extended agent frameworks with custom support for handoffs and interoperability (A2A protocol)
• Implemented observability and tracing for LLM and agent-based systems in production environments
• Integrated multiple LLM providers (OpenAI, Anthropic, Gemini) across different use cases

Development of a production-grade MLOps platform enabling reliable deployment, serving, and lifecycle management of Machine Learning models within a large-scale enterprise PaaS environment (e-commerce), supporting high-throughput, low-latency inference workloads across multiple customer-facing channels worldwide.

• Designed scalable model serving solutions supporting TensorFlow, ONNX, Scikit-learn, and other frameworks
• Implemented REST and gRPC inference endpoints compliant with V2 Inference Protocol
• Optimized deployments on OpenShift for latency, concurrency, and resource efficiency (load testing and performance tuning)
• Built CI/CD workflows for automated model promotion using GitHub Actions and Azure Machine Learning
• Implemented monitoring and observability using Prometheus and Grafana for model serving workloads

Worked on European-funded research projects (FlexiGroBots, Grapevine), focusing on ML infrastructure, workflow orchestration, and cloud-native systems.

• Managed Kubernetes clusters and multi-user Kubeflow platform for ML experimentation
• Developed Kubeflow pipelines supporting model serving and ML lifecycle workflows
• Contributed to a Cloudify plugin for cross-platform (HPC + Cloud) workload orchestration
• Coordinated technical work across project partners and contributed to project reviews

DevOps-focused software development for 5G Core network systems, contributing to the deployment and reliability of cloud-native microservices in a telecom environment.

• Deployed and managed microservices in distributed environments using Docker and Kubernetes
• Developed and maintained CI/CD pipelines and development infrastructure
• Automated build, deployment, and release processes
• Validated integration of third-party components ensuring system stability

Applied AI development across speech, computer vision, and conversational systems, focusing on real-world integration and early-stage production feasibility.

• Developed speech and speaker recognition systems using embedding-based models and performance evaluation
• Designed a custom vector similarity search layer (cosine similarity) on top of MongoDB
• Built real-time pipelines for processing large-scale telephone audio data
• Integrated speech-to-text and text-to-speech services with intent-based conversational systems (RASA)
• Developed conversational AI prototypes for public administration services

Applied research in additive manufacturing combining computational modelling, simulation, and data infrastructure.

• Designed and operated distributed data and compute infrastructure (Hadoop and Linux clusters)
• Performed multi-scale simulations of additive manufacturing processes
• Conducted molecular dynamics studies on graphene nanocoatings for steel
• Prototyped an electromagnetic atomization system for steel powder production
• Contributed to R&D projects (~€180k), including planning and technical execution

Supported research activities in additive manufacturing, contributing to simulation tasks and technical documentation within industrial R&D projects.

Carried out computational (Monte Carlo Dynamics) study of neutron-radiation damage in materials designed for future nuclear fusion reactors.