Adrián Bulnes

Infrastructure & Platform Engineering

Kubernetes, AWS, GPU/AI Infrastructure | ex-Intel, Dolby, Anomalo

Hello, my name is Adrian. I am an infrastructure and platform engineer with 14 years across cloud, embedded systems and silicon. At Anomalo I built and ran Kubernetes platforms inside enterprise customers' own clusters; before that, AWS micro-service architecture at Dolby, and embedded software and firmware leadership at Intel. Since 2024 I have led engineering at two ventures I co-founded, where AI meets custom hardware. I have worked fully remote since 2019.

During my free time I keep myself active with different sports, paragliding, caving, canyoneering, riding my motorcycle, or geeking at home.

  • LinkedIn

Over the last 14 years I have specialized in infrastructure, cloud platforms and embedded systems, through university degrees, work experience and self-learning. My core is Kubernetes and AWS, with Terraform, Helm, ArgoCD and CI/CD around them. What I am best at is making a platform run inside an environment I do not own, such as a customer's own cluster. More recently that has extended to GPU and AI infrastructure, and back down to the silicon it runs on.

Kubernetes, Helm, ArgoCD, Terraform, Docker, Linux, CI/CD

AWS: EKS, Lambda, API Gateway, CloudFormation, Cognito, DynamoDB, SQS, ECS/Fargate

Python, C#, C, JavaScript, Bash, Git

NVIDIA CUDA, PyTorch, GPU-resident inference

Instruction-set and microarchitecture design, I2C/SPI/UART/PCIe, GPIB test automation

I take a small number of independent projects a year, all of them the same shape: getting AI and Kubernetes workloads running in an environment you do not control — your customer’s own cluster, on-premises, at the edge, or air-gapped. It is what I did at Anomalo, where the product had to install and upgrade inside enterprise customers’ clusters, and it is the part most teams have never had to build.

Each engagement is a defined deliverable for a fixed fee. I sell these three as packages, and anything adjacent is worth asking about.

1–2 weeks

Deployability assessment

Can your product actually be installed in a customer’s cluster, on-premises or air-gapped? I go through it and tell you what breaks, in a written report with a plan and an estimate to fix it.

4–8 weeks

Customer-environment install kit

Helm charts, Terraform and an offline bundle that install your product in someone else’s Kubernetes, with upgrade and rollback paths and documentation their operations team can follow. Delivered against milestones.

2–4 weeks

Self-hosted LLM inference stand-up

Models serving on your own GPUs instead of an API: Kubernetes with the NVIDIA GPU Operator, vLLM or llama.cpp, monitoring, and a handover your team can run without me.

2025 - Present

Sparsr

Systems / Infrastructure Lead & Co-founder

Co-founding a hardware startup building a processor for binary matrix computation with 8,192-bit native registers, on an architecture I first designed and prototyped in 2019-2021: instruction-set design, multi-cycle to pipelined microarchitecture, and the simulation and verification toolchain. I own the compute and developer infrastructure end to end, from a bare-metal Linux GPU environment (NVIDIA CUDA, PyTorch, GPU-resident model inference) to self-hosted CI runners that run automated hardware health verification on every build.

2024 - 2025

ScaleCortex, Inc.

Lead Systems Engineer & Co-founder

Designed and built an agentic AI platform specialized in edge deployments: agents, self-hosted LLMs and GPU inference running in Kubernetes on our own bare-metal cluster, configured through a web interface that laid the agents out as a component diagram. Owned systems architecture and deployment tooling, alongside team coordination and business development. Company wound down in 2025.

2022 - 2024

Anomalo, Inc.

Sr. Cloud Infrastructure Engineer

Delivered and operated Kubernetes deployments of a data-quality platform inside enterprise customers' own clusters, using Helm, ArgoCD, AWS EKS, Terraform, Docker Compose, Python, Git and Linux. Worked directly with customer platform teams to make the product install, upgrade and run reliably across heterogeneous and access-restricted environments. Ran monitoring and observability and was on call, including incident response. Led the SOC 2 Type II compliance workstreams for infrastructure security and access controls, through two annual audits, both passed.

2019 - 2022

Dolby Laboratories

Staff Engineer

Built a cloud platform for testing Dolby's audio and video codecs: a web portal that launched test runs on cloud infrastructure and ran them in parallel, designed for customers to test the codecs with their own products. 6 micro-services, database, REST APIs, 2 clients and 2 long-running services, provider-independent, using AWS Lambda, API Gateway, CloudFormation, Cognito, DynamoDB, SQS, Docker, ECS+Fargate, C# and JavaScript. Line manager for 3 engineers in a 30-person organization, working across several teams at once.

2018 - 2019

Intel Corporation

Software Architect

Design of all software components for a custom chip used in datacenters, working side to side with the other architects (power electronics, digital electronics, chip packaging).

2016 - 2018

Intel Corporation

Software Technical Lead

Responsible for the entire Firmware Validation of a next-gen communications device used in data centers.

2013 - 2016

Intel Corporation

Firmware Validation Engineer

Development of the Firmware Test Framework and Test Cases of Intel Atom camera co-processor. Rebuilt the test platform on my own. The old harness was a Windows GUI that someone had to drive by hand, and a sequential run took hours; the new one ran unattended and in parallel on Linux servers across a data center, with none of the thousands of existing test cases changed. Intel gave it an award for improving validation throughput by 10x.

2012

European Space Agency

Software Engineer (Master's final placement)

Sole developer of the control software and hardware integration for ESA's Videogrammetry System, a computer-controlled electromechanical device that measures spacecraft deformation inside the Large Space Simulator thermal-vacuum chamber, and is still in use at ESTEC. The old software only drove some of the components; I rewrote it from scratch for the updated ones and integrated new hardware, delivered unsupervised, interfacing with precision, space-grade instrumentation via C, C# and USB.

2015

Intel Recognition Award

Cross-team collaboration. On my own time, I instrumented the firmware team's test runs so their developers could read the results faster, with exported metrics and charts of how the product behaved under test.

2014

Intel Award

“Improving validation throughput by 10x”

Feb 2011

"Training with microcontrollers using Curuxa"

University of Seville

May 2010

"Curuxa"

University of Cadiz

May 2008

"Pigmeo"

University of Seville