I'm an AI architect at Oracle's AI Centre of Excellence in Spain. Most of my time goes into building and benchmarking GPU-accelerated AI and HPC infrastructure on the NVIDIA AI Enterprise stack. The rest goes into open-source governed-AI projects, and writing up what I learn along the way.
About
I've worked on high-performance computing and AI infrastructure for a little over twenty years. That has taken me from automotive CAE clusters and financial-services HPC to GPU-accelerated GenAI platforms on the cloud, across a lot of different industries and teams.
These days my focus is reference architectures, performance baselines, and the guardrails that keep GPU GenAI and HPC workloads dependable. When I step away from customer work, I usually end up building an open-source framework, running a benchmark, or writing about what actually worked and what didn't.
Expertise
I design, deploy, and benchmark large AI and HPC environments, and I like to stay hands-on. Most of it runs on the NVIDIA AI Enterprise stack.
- NVIDIA AI Enterprise: NIM, NeMo, Triton, TensorRT, TAO, RAPIDS / cuVS, DeepStream, cuOpt
- GPU infrastructure: large A100, H100 and H200 clusters, multi-node GenAI and distributed LLM training
- Benchmarking and tuning: throughput, latency, scaling, and cost per token across inference and training
- Cloud and platform: OCI, AWS, GCP and Azure, plus Kubernetes, Slurm, Terraform and Ansible
Experience
Selected open source
pip install antahkarana (plus antahkarana-cli); weights are on Hugging Face as antahkarana-base, and there are two arXiv papers.Publications
Book
Antaḥkaraṇa: The Inner Instrument
I took a 2,500-year-old map of the mind, the antaḥkaraṇa or "inner instrument," and read it as an engineer. The book is the story of building it, one organ at a time, into a real, measured continual-learning AI, with the failures left in and every number you can check.
Beyond technology
Away from infrastructure, I recharge with cricket and music. Cricket has been a lifelong thing for me. It is all patience and strategy, and the balance between doing your own job well and playing for the team, which honestly is not so different from how good infrastructure work feels.
Music is the other side of my head. Singing and listening across a lot of genres is my counterweight to the technical work, and where I go to reset.
Contact
Happy to talk about AI and HPC infrastructure, GPU platforms, or governed AI. LinkedIn is the easiest way to reach me.