ABOUT
Demand intelligence
that actually runs in production.
Nalu AI is an ML system built together with the customer — for demand planning and sales intelligence, on their infrastructure, for their data.
WHAT NALU AI IS
Not SaaS. Not a consulting project.
Nalu AI is a demand intelligence platform tailored to each customer. Forecasts, SCM recommendations and AI reports run on the customer's own infrastructure — no shared data, no external cloud, no black box. The platform is built on proven technology from production use: LightGBM and Temporal Fusion Transformer for forecasts, DuckDB for analytical queries, FastAPI as the API layer, Docker for deployment. What works stays — what matters in mid-market gets added.
Not generic.
Every implementation learns the customer's SKUs, seasonalities and ERP quirks — no standard templates.
Not dependent.
On-premise by design. No cloud lock-in, no subprocessor chain, no hidden APIs.
Not bloated.
One Docker Compose stack per customer. What is not needed does not run. What runs, runs fast.
Nalu ist Hawaiianisch für Welle — entstanden während eines Sabbaticals auf Hawaii, wo nebenbei ein Dive-Shop-System entwickelt wurde.
METHOD
How a project runs.
From the first conversation to production — typically a few weeks, depending on data quality and ERP complexity.
- 01
Discovery
Short workshop with the customer: ERP system, data structure, business logic, biggest pain points. Out of it comes a config.yaml — the heart of every implementation.
- 02
Connector & Data pipeline
Integration with SAP, S/4HANA, Dynamics, Proalpha, SQL or CSV — read-only, no changes to the source system. Data quality is validated automatically, gaps detected.
- 03
Model training
ML models are trained on the customer's real data. Forecast accuracy is transparently measured (MAPE, coverage) before anything goes into production.
- 04
Deployment
Docker stack runs on the customer's server. RBAC, 2FA, audit log and SSL are standard. Training for purchasing, planning and sales.
- 05
Operations & maintenance
Weekly retraining, automatic reports, monitoring. Updates, new features and support are part of the license — not an upcharge.
TECHNOLOGIE
PRINCIPLES
Four things Nalu AI measures itself against.
Data stays with the customer
On-premise is not a feature, it is a requirement. No cloud dependency, no subprocessor chain, no external model APIs.
No hype, just results
Forecasts, alerts and reports the team can use tomorrow. No generic AI buzzwords — confidence intervals, SHAP explanations, honest numbers.
Maintainable over clever
Deterministic pipelines, documented configuration, readable models. Whoever has to touch the system in three years should not have to guess.
Mid-market understands mid-market
Tight resources, legacy systems, pragmatic decisions. The platform is built for that — not for enterprises with their own data team.
Learn more.
No form. No funnel. Just an email.
aloha@nalu-ai.comTypical investment: from €50,000 one-time, from €5,000/month — always individually calculated.