Professional Portfolio
MLOps / LLMOps Engineer · PhD Researcher in Electrical and Computer Engineering · AI Systems Engineer
Production AI, intelligent Cyber-Physical Systems, Digital Twins, predictive maintenance, Explainable AI and human-centred industrial intelligence.
About
I am a Computer Engineer whose career evolved from computer networks, databases and institutional IT infrastructure into software engineering, cloud-native platforms, Machine Learning, MLOps, LLMOps and intelligent industrial systems.
I currently work at LynxMind Portugal in DevOps, MLOps and LLMOps-oriented engineering, designing, automating and monitoring data and Machine Learning pipelines in production environments. My work includes Linux-based systems, CI/CD, Docker, Kubernetes, MLflow, DVC, Apache Airflow, model operationalization, cloud-native deployment and security-aware delivery.
In parallel, I am pursuing a PhD in Electrical and Computer Engineering at FEUP – University of Porto and carry out research within INESC TEC / LIAAD. My doctoral work investigates intelligent and explainable Cyber-Physical Systems for autonomous predictive maintenance in distributed Industry 4.0 and Industry 5.0 environments.
The research combines unsupervised anomaly detection and degradation monitoring, Digital Twins, Explainable AI, uncertainty communication and human-in-the-loop decision support. Large Language Models are explored as an interaction layer that helps maintenance professionals query validated system outputs, Digital Twin states and explanation evidence using natural language.
Career & Academic Journey
A progression built across public education, academic networking, university teaching, international training, cloud engineering, AI systems and doctoral research.
Began formal technical training in computing with a focus on computer networks and databases, establishing the infrastructure foundations that later supported work in systems administration, distributed computing and AI platforms.
Developed academic foundations in information technology while progressively combining university studies with professional responsibilities in educational planning, information management and technology infrastructure.
Worked in district education planning while completing the undergraduate degree, developing experience in institutional planning, information organization and public-service operations.
Accepted a transfer to the agricultural institute, where I taught computing and administered the institution's Internet and network infrastructure.
Took responsibility as a focal point for MoRENet, contributing to academic network connectivity and infrastructure coordination. During the same period, I completed specialized practical training in TCP/IP networking protocols.
Joined Universidade Licungo as a lecturer while continuing responsibilities linked to MoRENet. Worked with the university-hosted data-centre infrastructure supporting Internet distribution across central and northern regions of Mozambique.
Advanced from systems and infrastructure toward applied Machine Learning and production-oriented data systems. The degree included an international Machine Learning training period in Barcelona focused on translating analytical models and data pipelines into practical solutions.
Six-month training linked to the Master's program, focused on applied Machine Learning, business-oriented data analysis, model development and communicating analytical outcomes to different stakeholders.
Design, automation and monitoring of Machine Learning pipelines in production environments, combining CI/CD, container orchestration, experiment tracking, model deployment, data governance, observability and DevSecOps practices.
Experience configuring cloud security and vulnerability-management tooling in Linux environments, designing network routes between environments, working with Docker and Kubernetes, and supporting secure, highly available infrastructure.
Completed postgraduate specialization in Generative Artificial Intelligence at FCUP, covering foundation models, LLMs, RAG, fine-tuning and LLMOps. Also obtained Certified Ethical Hacker training/certification and contributed to scientific peer-review activity.
Doctoral research on intelligent and explainable Cyber-Physical Systems for autonomous predictive maintenance in distributed Industry 4.0 and Industry 5.0, integrating unsupervised learning, Digital Twins, XAI and human-in-the-loop decision support.
Specialized training covering industrial automation, edge computing, IoT security, system integration and human-centred solution design.
Services
Reproducible ML/LLM pipelines, governed model access, RAG and agentic workflows, evaluation, orchestration, model serving, CI/CD and production lifecycle management.
Cloud-native platforms for AI and distributed workloads using container orchestration, scalable infrastructure, edge/cloud integration and production-ready service design.
Automated build, validation, deployment and release workflows for software and AI systems, with traceability, reliability and security integrated into delivery.
Data-driven modelling, deep learning, Computer Vision, model evaluation and Explainable AI for research and production-oriented applications.
AIOps-assisted incident analysis, root-cause investigation, model and platform telemetry, drift detection, performance tracking and security-aware delivery for production systems.
Intelligent CPS architectures connecting sensors, edge/cloud processing, predictive models, Digital Twins, XAI and human decision support for Industry 4.0/5.0.
Technology Stack
Doctoral Research
Research in distributed Industry 4.0 and Industry 5.0 ecosystems, integrating predictive AI, Digital Twins, continuous explainability and human-centred decision support.
Sensors, industrial protocols, operational context and distributed data acquisition.
Unsupervised anomaly detection, degradation monitoring, robustness and concept drift.
Asset state, synchronization, degradation signals and maintenance-oriented context.
Feature evidence, causal traces, confidence, uncertainty and explanation auditability.
Natural-language access to validated system outputs, Twin states and explanation evidence.
Operator validation, feedback, override, maintenance planning and decision confidence.
Main focus on unsupervised learning for normality modelling and anomaly/degradation detection, with autoencoders as the primary candidate and comparison with temporal and classical anomaly-detection approaches.
Continuous explainability across requirements, modelling, monitoring and use, evaluating fidelity, stability, clarity, actionability, uncertainty communication and decision usefulness.
Credentials & Recognition
Promoted by the World Bank and Nigeria's Ministry of Technology.
International recognition in a Huawei technology program in China focused on IoT and Computer Vision.
Postgraduate Specialization · FCUP, University of Porto · 2025
Cybersecurity, vulnerability assessment and penetration-testing foundations · 2025
Industrial automation, edge computing, IoT security and Design Thinking · 2026
Advanced neural-network and deep-learning training.
Cloud-based Machine Learning and AI platform training.
Cloud infrastructure and Azure platform training.
Selected Projects
Selected work includes two enterprise AIOps projects from the attached repositories, alongside representative GenAI, MLOps and DevOps engineering projects. Descriptions focus on verified architecture and engineering outcomes without inventing unsupported performance metrics.
PARDINUS · AGENTIC RCA · CHECKMK
AI-assisted Root Cause Analysis and proactive operations extension for Checkmk 2.5+. It ranks evidence across alerts, logs, topology and service history, reconstructs incident timelines, calculates confidence, and enriches investigations with deep host telemetry and distributed-monitoring context.
The platform also includes proactive RCA with storm and rate-limit protection, a floating operator chatbot, specialist agents, bounded ReAct investigations, permission-gated domain diagnostics, business-impact context, token usage tracking, and offline licensing/code-protection controls.
Engineering outcome: Combines reactive and proactive incident investigation, governed AI assistance and controlled diagnostics in a deployable Checkmk MKP while preserving auditability and deterministic fallback when the LLM provider is unavailable.
AI-TOWER · LAIOP · SOVEREIGN AIOPS
Architecture and monorepo for a sovereign AIOps platform that ingests operational signals, validates and redacts sensitive data, streams durable events, and orchestrates the incident lifecycle from correlation and triage through RCA, recommendation, human approval, remediation, validation, rollback and reporting.
The design supports SaaS, customer-managed VPC and on-premises deployment, BYO-LLM/BYOK model sovereignty, retrieval-grounded reasoning, governed AI access and full platform observability. The attached snapshot includes a runnable FastAPI backend foundation plus modular application and package scaffolding for the wider architecture.
Engineering outcome: Establishes a cloud-agnostic, event-driven foundation for trustworthy incident automation with data sovereignty, human-in-the-loop controls, graceful degradation and end-to-end observability as architectural requirements.
Research-oriented LLM interface that converts validated predictive-maintenance outputs, Digital Twin state and XAI evidence into contextual natural-language queries, summaries and maintenance recommendations.
Outcome: Human-machine interaction architecture aligned with explainability and decision support rather than unconstrained text generation.
Modular architecture for document ingestion, retrieval, prompt routing, tool use and response evaluation, designed to move GenAI prototypes toward controlled production workflows.
Outcome: Clear separation between ingestion, retrieval, orchestration, evaluation and serving layers for maintainable GenAI systems.
Reproducible pipeline for data preparation, experiment tracking, unsupervised model training, validation, packaging and deployment in predictive-maintenance scenarios with temporal sensor data.
Outcome: Traceable lifecycle from data and experiments to model serving and operational monitoring.
Monitoring layer for model performance, drift signals, inference behaviour and operational service health, connecting ML evaluation with platform-level observability.
Outcome: Unified visibility for model behaviour and infrastructure conditions across the production lifecycle.
Automated delivery workflow for software and AI services, including build, test, containerization, deployment promotion and environment-specific release controls.
Outcome: Repeatable deployments with stronger traceability and reduced dependence on manual release procedures.
Security-aware container delivery and runtime operations combining CI/CD controls, Linux hardening, vulnerability-management practices and Kubernetes operational safeguards.
Outcome: Delivery practices aligned with availability, compliance and security requirements.
Operational automation spanning Linux and Windows systems, virtualization, network connectivity, remote administration and containerized services across hybrid environments.
Outcome: Infrastructure experience connecting traditional networks and systems administration with modern cloud-native operations.
Contact