Every resource belongs to one Space
Datasets, connectors, designs, runs and inference servers live inside a single Space, and access is granted per Space through seats — so an administrator in one Space has nothing in another. Spaces
The Large Data Model (LDM)
An LDM learns across your structured and unstructured data in one model, instead of one pipeline per question. Below is what that means once it is a product: the guarantee each stage gives you, and where the documentation covers it.
One path from connected data to a served model
Read your sources where they are
Connect warehouses, lakes, databases and APIs as live integrations. NeoData reads through them directly — there is no ETL pipeline to build and no export to keep in sync. Data Integration
Runs start from a published design
A Model Design fixes the datasets, the time split, the targets and the metrics before any run exists. Two runs from the same design are comparable, and every run points back to what it was built from. Model Design
Pre-train once, fine-tune per task
Pre-training builds foundation checkpoints over your data; fine-tuning adapts one of them to a specific task. Each has its own runs and leaderboard, so a new question does not mean training from scratch. Pre-training
Leaderboards rank checkpoints, not runs
A run’s best checkpoint is rarely its last. Benchmarks hold the comparison fixed, so an improvement you measure is an improvement you can ship. Evaluation
Online, batch, and watched for drift
Deploy a checkpoint as an HTTPS endpoint for live requests, or score a whole population offline in one pass. Drift Radar compares each target against a baseline and flags what has moved. Inference Server
Read the full LDM introduction in the documentation →
How it works
A single path from connected data to production inference — modeled after how teams ship AI on NeoData.
Connect your data sources
Lakes, warehouses, streams, APIs
Build dataset
Select paths, validate schema
Model design
Datasets, split, targets, metrics
Pre-train
Foundation on your corpus
Fine-tuning
Adapt a checkpoint to one task
Inference
Score at production scale
Land structured and unstructured data in NeoData with governed pipelines. The LDM only sees what you connect and approve — ready for training and audit.
After the workflow
Ship LDM with the same rigor you use for the rest of NeoData
The steps above are the product story. Next, these docs cover Spaces, APIs, permissions, and drift monitoring — so what you validate in pre-production still holds when traffic, data drift, and compliance requirements show up for real.
What partners achieve with NeoSpace
Representative outcomes when LDM is deployed for high-stakes decisioning at scale.
2×
Prediction Accuracy
More accurate predictions powered by our advanced data modeling approach.
400%
Training Speed
Faster model training cycles, accelerating experimentation and deployment.
5×
Inference Latency
Faster inference speeds deliver instant responses across massive datasets.
10B+
Real-Time Scale
Predictions processed in real time, scaling effortlessly to billions of events.
Ready to unlock the value hidden in your largest data assets?
Join leading companies already using NeoSpace LDM to transform data into actionable intelligence.