About
Independent data and AI infrastructure expert. Performance, scalability, reliability, resilience and cost, reviewed across the whole stack.
I am Julien Laurenceau. I work with CTOs, CIOs, platform and data leaders, SRE teams and investors on the questions that decide how a data or AI platform performs, what it costs, and whether it holds when something fails.
Organizations bring me in at two moments. Before a commitment: a scale-up, a migration, an architecture redesign, a hardware or vendor purchase, a transaction. And when a platform in production stops behaving predictably and the internal diagnosis has run out of road. In both cases the value is the same: an independent technical read, delivered fast, that someone can act on.
I do not stop at the layer boundary
My background is unusual in one specific way: I am comfortable across the whole stack rather than inside one product. In a single engagement I might read application source code, review a data model, profile a distributed processing engine, inspect Kubernetes scheduling and network paths, analyze the behavior of a Ceph or MinIO cluster, check operating system and infrastructure configuration, and then compare the drive specifications in a vendor quote against what those drives deliver under the access pattern the workload actually produces.
That range comes from twenty years spent on both sides of the line, in HPC and numerical optimization first, then in production data infrastructure. It is what lets me say where a problem is rather than where it appears, and it is the reason my sizing recommendations tend to be smaller than the ones vendors propose.
Background
PhD in numerical optimization, with ten years in high performance computing. More than twenty years of hands-on production experience in distributed systems, data platforms and private cloud.
I have worked in banking, telecommunications, aerospace and satellite imagery, on systems where performance and reliability have direct financial consequences: multi-petabyte data lakes, Spark on Kubernetes, Kafka pipelines, Ceph and MinIO object storage, Cassandra and ScyllaDB, PostgreSQL, and the infrastructure underneath AI training and inference.
Where I’ve gone deep
- Multi-petabyte data lakes on private cloud
- Spark on Kubernetes: performance, reliability, cost
- Kafka pipelines: sizing, performance, reliability
- Object storage at scale: Ceph, MinIO, S3
- Cassandra and ScyllaDB in production / banking contexts
- Resilience, disaster recovery and RTO / RPO validation
- Capacity planning and hardware sizing, from workload to server and drive selection
- Satellite imagery and geospatial pipelines
- HPC and numerical optimization
- AI infrastructure and training data pipelines
- Private cloud architecture and capacity planning
How I work
Remote-first, scoped engagements with clear deliverables. I prefer short, high-leverage work (audits, due diligence, capacity planning, flash reviews, expert calls, fractional advisory) over long staff-augmentation missions. The goal is to leave your team with a sharper picture of their own system and a prioritized roadmap, not a dependency on me.
I sell no implementation, no licenses and no hardware, and I have no partnership with any vendor. Nothing in my conclusions depends on what you buy next. When my recommendation is to buy less, spend nothing or keep what you have, there is no incentive pulling the other way.
Based in Bordeaux, France. I work in English and French, with clients across Europe and North America, and I schedule interviews and debriefs around the client’s time zone.