Description
ModelMill is a cloud-based deep-learning platform developed by Literal Labs for training, testing, and deploying Logic-Based Networks (LBNs). It is designed to automate the end-to-end process of AI model creation, from data import to SDK generation, allowing users to produce edge-ready AI models without needing specialist AI engineering expertise.
The platform accepts datasets in CSV, JSON, or ZIP formats, up to 5GB in size, and includes curated OpenML tabular datasets for those without their own data. ModelMill pre-processes data automatically, performing operations like filtering, normalization, and sequence padding to prepare it for high-quality training.
Annotation is a key step in ModelMill's workflow. Users can annotate data manually or leverage ModelMill's AutoML engine, which proposes annotations based on statistical analysis and structural patterns. The platform supports classification, regression, and anomaly detection tasks, automatically detecting the task type based on data annotations.
Once data is annotated, users define the deployment target, including hardware constraints and business metrics. ModelMill then trains hundreds or thousands of LBN configurations in parallel, refining them to produce a shortlist of candidates that meet deployment requirements.
The final step involves converting the chosen model into a C code SDK, ready for integration into embedded or server codebases. The SDK includes the trained LBN, inference engine, build configuration, example code, and documentation, and is tested across various processor architectures, including ARM, ESP, RISC-V, and x86.
ModelMill offers several pricing plans, including a Demo plan for evaluation, with limited training runs and dataset sizes. The platform is designed for technically capable teams without dedicated AI staff, automating configuration, training, and candidate selection.
ModelMill's Core Features
Automated deep-learning platform
Logic-Based Networks training
Data import and pre-processing
AutoML-assisted annotation
Parallel model training
C code SDK generation
Edge hardware deployment
OpenML dataset integration
No GPU requirement
Supports CSV, JSON, ZIP formats
Classification, regression, anomaly detection
Continuous performance evaluation
Deployment royalty model
Custom benchmarking visualizations
Team collaboration features
How to use ModelMill?
Upload data: Import datasets in CSV, JSON, or ZIP formats
Annotate: Use AutoML engine for smart annotations
Define deployment: Specify hardware and business metrics
Train models: Parallel training of LBN configurations
Select candidates: Evaluate and choose best-performing models
Generate SDK: Convert model to C code for integration
Deploy: Integrate SDK into edge hardware
ModelMill's Use Cases
- Edge AI deployment
- Data annotation
- Parallel model training
- SDK generation
- OpenML dataset usage
