AI Systems That Make Work Faster, Smarter, Human.

I design, build, and ship production AI — RAG pipelines, fine-tuned LLMs, and agentic workflows for organisations that need answers, not demos.

huberty.ai / console

Retrieval dashboard

Welcome back, Nicolas

DOCUMENTS INDEXED

1.24M

across 9 sources

MEAN LATENCY

240ms

p95 retrieval

ANSWER ACCURACY

96.4%

vs. graded set

How I think

I'm an engineer at heart and a builder by obsession.

Two UCLouvain degrees and five years shipping production ML, RAG, and agentic workflows  currently leading the AI practice at UCLouvain, previously building national-scale data infrastructure at Smals. I believe AI should make professionals faster, not replace them. Technology should empower, not overwhelm.

5+

Years in production ML

12+

Systems shipped

How I work
How I work

From raw idea to shipped system in weeks, not quarters.

01

Discover

Map the real problem

Your data, your workflows, your constraints. I sit with the people doing the work before writing a line of code.

02

Build

Ship intelligence that holds up

RAG pipelines, fine-tuned LLMs, and agentic workflows — engineered for accuracy, latency, and explainability, not demos.

03

Deploy

Run it in production

Kubernetes, CI/CD, and monitoring on private infrastructure. It works on day one and keeps working at scale.

The toolkit

One stack, end to end.

From model training to the Kubernetes cluster it runs on — I own the whole pipeline so nothing falls between the cracks.

AIDataWebCloud
PythonPyTorchLangChainLangGraphHugging FaceOpenAIPostgreSQLQdrantPythonPyTorchLangChainLangGraphHugging FaceOpenAIPostgreSQLQdrant
RedisKafkaAirflowdbtTypeScriptNext.jsNuxtFastAPIRedisKafkaAirflowdbtTypeScriptNext.jsNuxtFastAPI
ReactTailwindDockerKubernetesAzureAWSTerraformGitHub ActionsReactTailwindDockerKubernetesAzureAWSTerraformGitHub Actions