Our stack at a glance
Selection criteria: adopted in production (not just tested), used for at least 3 months, measurable value on client projects. We only recommend what we actually use.
01
Cursor
AI IDE · Development
The code editor that changed our development pace. Contextual autocomplete and integrated chat have halved our prototyping time.
02
LangSmith
Observability · LLM
The reference platform for monitoring, debugging, and evaluating LLM pipelines. Essential in production. Tracing, evaluation, datasets : it's all there.
03
dbt Cloud
Data Engineering · Transformation
The standard for data transformation. Versioned, tested, documented. We never run a data project without dbt. Automatic documentation is a game-changer for clients.
04
Claude (API)
LLM · Generation
Our go-to LLM for complex reasoning, extraction and code generation tasks. The 200k token context window is a real differentiator on document-heavy projects.
05
Dagster
Orchestration · Data
The data orchestrator that replaced Airflow in our new projects. The asset-centric model makes pipelines far more readable and maintainable long-term.
06
Weaviate
Vector Database · RAG
For projects requiring a dedicated vector store at scale. Multimodal, multi-tenant, with excellent hybrid search performance.
07
Weights & Biases
MLOps · Experimentation
The ML experiment tracker that has become our collective memory. Every run logged, every hyperparameter tracked. Essential for fine-tuning projects.
08
Retool
No-code · Internal Tools
For quickly building internal interfaces connected to our AI pipelines. 10× faster than a custom app for admin tools or human-in-the-loop validation.
09
Grafana + Prometheus
Monitoring · Infrastructure
Our monitoring stack for everything in production. Latency, errors, inference costs : all visualized. We deploy nothing without a Grafana dashboard.
10
Notion AI
Knowledge Management · AI
For internal documentation and project knowledge management. The integrated AI streamlines meeting summaries, spec writing, and knowledge base searches.
What's not on this list: dozens of tools tested and abandoned. The noise in the AI ecosystem is immense. Our filter: are we still using it 3 months after discovering it?
Tools to watch in 2026
Without over-anticipating, a few tools are maturing and could join our stack this year: Modal for serverless ML model deployment, LlamaIndex Workflows for agentic orchestration, and Evidence for BI as code.
Going further
Assembling these tools into one coherent workflow, without coding, is exactly what the AI Business Accelerator does.
Sylvie Wendkuni NITIEMA
Founder & Data Scientist · DataSAI
Over ten years in AI and data consulting (Big 4, telecoms, finance). She helps companies deploy AI in production and leads the DataSAI Academy.