Adrián Fernández
de la Torre

Platform Architect. Open Source Contributor. Cloud-native systems design with a focus on secure, scalable developer platforms.

Don't believe in AI hype (it's a just powerful tool as any other), compilers were once considered magic too.

PortfolioExperience

Portfolio

Research

Modeling Emergent Vegetation Dynamics: An ARC-NCA Framework

Reconstructing vegetation indicators from RGB drone imagery using Neural Cellular Automata.

This research project explores how Neural Cellular Automata can reconstruct vegetation information from standard RGB drone imagery. Developed as part of my Master’s thesis, the project uses an ARC-NCA model to estimate near-infrared reflectance and derive NDVI maps, reducing the reliance on specialised multispectral sensors.

Unlike conventional image-to-image models, the ARC-NCA evolves its predictions through repeated local interactions between neighbouring cells. This self-organising process enables the model to progressively reconstruct spatial vegetation patterns while offering a more interpretable perspective on how those patterns emerge.

The project includes an end-to-end geospatial deep-learning pipeline for processing UAV orthomosaics, aligning multispectral data, extracting image patches, integrating spatial information and agricultural annotations, training the neural cellular automaton, and evaluating its predictions using metrics such as RMSE, MAE, R² and SSIM.

By combining remote sensing, deep learning and emergent computation, the project investigates a novel and potentially accessible approach to crop-health monitoring and vegetation analysis in precision agriculture.

PythonAINeural NetworksRemote SensingGenerative ModelDeep Learning

Deep Learning for Earthquake Damage Detection

Accelerating satellite-based earthquake damage classification through deep learning and edge inference.

This project explores the use of deep learning and high-performance inference to accelerate earthquake damage assessment from satellite imagery. Originally developed as part of my Bachelor’s degree project, it uses a convolutional neural network to classify areas captured in Very High Resolution imagery as either damaged or undamaged.

The model was trained using GeoEye-1 satellite images collected after the 2010 Haiti earthquake. Built with Python, TensorFlow and Keras, the network learns visual patterns associated with structural damage and provides an automated mechanism for analysing affected areas—supporting faster decision-making during time-critical disaster-response operations.

Beyond developing the classification model, the project focuses on making deep-learning inference portable and efficient across different hardware environments. Using the Intel OpenVINO toolkit, the trained network was converted, optimised and deployed on Intel CPUs and the Intel Movidius Neural Compute Stick 2. The implementation evaluates synchronous and asynchronous inference, heterogeneous execution and mixed-precision optimisation, while comparing OpenVINO performance with TensorFlow Serving.

The research demonstrated that hardware-aware optimisation can significantly reduce inference time while maintaining reliable classification accuracy. This makes the solution suitable for edge and resource-constrained environments, where satellite images may need to be processed quickly without relying entirely on large cloud-computing infrastructure.

The work resulted in the peer-reviewed publication “Portability and Acceleration of Deep Learning Inferences to Detect Rapid Earthquake Damage From VHR Remote Sensing Images Using Intel OpenVINO Toolkit,” published in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

By combining remote sensing, convolutional neural networks and edge-AI optimisation, the project demonstrates how portable deep-learning systems can support faster and more effective post-disaster assessment.

PythonDeep LearningOpenVINORemote Sensing

Open-Source

Improving Flux with Distributed Tracing and Modern Events

Advancing Flux with end-to-end OpenTelemetry tracing and modern Kubernetes Events API support.

As an open-source contributor to the Flux project, I proposed and contributed to RFC-0011, introducing native OpenTelemetry tracing across Flux’s multi-controller GitOps workflows. The feature correlates source revisions, reconciliations and notifications into end-to-end distributed traces, helping operators understand how a change moves from a Git repository or OCI artifact through Kustomizations, HelmReleases and downstream notifications.

The design preserves Flux’s declarative and stateless architecture by extending its existing Alert and Provider APIs. Users can select the Flux resources they want to observe and export their traces to OpenTelemetry-compatible systems such as Jaeger or Grafana Tempo, without instrumenting individual controllers or deploying dedicated sidecars. RFC-0011 has been implemented and made generally available as part of Flux v2.7.0.

Alongside this work, I initiated an ongoing effort to migrate the Flux ecosystem from the legacy Kubernetes core/v1 event model to the modern events.k8s.io/v1 API. The initiative introduces an updated event-recording implementation while preserving compatibility with existing Flux notifications and integrations. It also coordinates changes across the shared Flux runtime and its source, Kustomize, Helm and notification controllers.

Together, these contributions improve Flux’s observability, troubleshooting capabilities and alignment with modern Kubernetes standards, making complex GitOps reconciliation workflows easier to trace, interpret and operate.

GoCNCFKubernetesFluxCDGitOpsOpenTelemetry

Experience

Platform Architect

EricssonNov 2025 – Present
Madrid, Spain

Leading architecture studies and strategic initiatives for a global AI-native development platform, with a particular focus on cloud-native development environments, CI/CD capabilities and modern software delivery practices. My role involves evaluating platform technologies, identifying opportunities for standardisation and defining architectural approaches that improve developer experience, scalability, security and operational consistency.

I also lead a team responsible for progressing these studies from initial exploration through technical assessment, recommendation and delivery planning. This includes coordinating activities, supporting/mentoring team members, reviewing technical outcomes and ensuring that proposed solutions align with wider platform objectives and organisational priorities.

A key part of the role is working with engineering teams, platform stakeholders and decision-makers across the organisation to articulate the architectural direction and build alignment around platform evolution. I contribute to roadmap definition and governance activities, helping establish clear priorities, technical principles and decision-making frameworks for the continued development of the global platform.

Through this work, I help shape the long-term direction of the organisation’s development and delivery ecosystem, balancing innovation with reliability, maintainability and the practical needs of engineering teams.

DevSecOps Engineer

EricssonSept 2024 – Nov 2025
Madrid, Spain

As part of a global innovation team, I contribute to the exploration, design and development of new ideas that can be incorporated into the company’s technology portfolio. My role involves assessing applications, platforms and services against cloud-native principles, helping teams improve areas such as automation, scalability, security, resilience and operational maintainability.

A key part of my work is promoting a DevOps mindset across engineering activities and supporting development teams in adopting collaborative, automated and security-conscious delivery practices. This includes encouraging shared ownership, continuous improvement and the integration of security considerations throughout the software development lifecycle.

I develop automation, integrations and supporting tools primarily in Python and Go, while working with additional technologies when required by the solution. I also help standardise deployment practices through GitOps principles, improving consistency, traceability and governance across application environments.

Alongside delivery and platform-engineering activities, I contribute to proof-of-concept initiatives and generative AI pilots, evaluating emerging technologies and transforming promising ideas into practical demonstrations for technical and business stakeholders.

I am also actively involved in open-source collaboration, contributing to CNCF projects and engaging with the wider cloud-native community. These contributions allow me to apply open-source engineering practices, collaborate with maintainers and bring relevant technologies and standards back into internal innovation initiatives.

DevOps Engineer

EricssonJune 2020 – Sept 2024
Madrid, Spain

As part of a globally distributed engineering team, I contributed to the design, development and delivery of cloud-native applications built around microservice architectures. Kubernetes served as the main platform for both development and deployment, while Helm, Helmfile and Python were used to package, configure and automate application delivery across different environments.

My role involved developing and maintaining CI/CD pipelines with GitLab CI/CD, Jenkins and Spinnaker, helping teams automate build, validation and deployment processes. I also supported the adoption of GitOps practices using Argo CD and Flux CD, assessing application and repository structures to improve deployment consistency, traceability and operational reliability.

Alongside delivery activities, I participated in feature-design and solution-architecture discussions, contributing implementation ideas and translating technical requirements into practical solutions. I developed reusable tools, scripts and deployment assets in Go and Python to automate recurring tasks and simplify engineering workflows.

I also supported the onboarding and development of new team members through mentoring, knowledge sharing and collaboration with teams across the organisation. Working within Agile and Scrum environments, I regularly contributed to planning sessions, reviews and retrospectives, while presenting technical designs, demonstrations and project progress to both the immediate team and wider internal audiences.

In addition, I contributed to generative AI pilots and proof-of-concept initiatives, exploring how emerging AI technologies could improve software delivery, automation and engineering productivity.

Implementation Engineer (Intern)

EricssonNov 2019 – June 2020
Madrid, Spain

As part of a globally distributed engineering team, I collaborated remotely with colleagues across different regions and time zones while supporting the design and implementation of automation solutions for customer environments. Under the guidance of experienced engineers, I contributed to different stages of the delivery process, including the development and maintenance of CI/CD pipelines for cloud-native applications.

I also participated in the deployment and troubleshooting of large-scale applications running on Kubernetes, gaining practical experience with containerized workloads, configuration management, observability and platform-related issues. Alongside these activities, I developed scripts, lightweight applications and internal tools to improve automation and simplify recurring engineering tasks.

The role also involved contributing to technical documentation, knowledge-sharing activities and the continuous improvement of implementation processes and reusable engineering practices.

IT Consultant (Intern)

EricssonMay 2019 – Sept 2019
Madrid, Spain

As an IT Consultant Intern, I supported the maintenance and evolution of a long-standing business application, working within an established codebase and contributing to improvements, issue resolution and ongoing functional development. This experience provided practical exposure to legacy systems, software lifecycle management and the challenges involved in extending mature applications while preserving stability and compatibility.

I also contributed to proof-of-concept initiatives exploring the feasibility of blockchain technologies for selected business use cases. This involved analysing requirements, assessing potential architectures and evaluating whether distributed-ledger solutions could provide meaningful technical or operational value.