Features
Comprehensive, execution-focused capabilities for AI and ML engineering, optimized for local and VS Code workflows.
Quick Overview
NEO accelerates AI and ML workflows with local-first autonomous agents:
Automate repetitive tasks and speed up your workflow
Multi-agent flows, RAG, tools/MCP, evals, and production LLM apps
Deploy to cloud, local, on-premises, or edge environments
Version models, datasets, and pipeline outputs
Agentic & LLM engineering
Neo is built for multi-step agents in VS Code: planning, tool use (including MCP), retrieval and multimodal RAG, multi-agent coordination, and shipping with evaluation and guardrails — all grounded in your repository and data.
Multi-agent & workflows
Spec-to-ship delivery, swarms, and specialized roles — Spec to Ship, Agent Swarm.
Tools & repository intelligence
Deep repo Q&A, MCP, and integrations — Repo Query Agent, MCP Optimization.
RAG & documents
Grounded apps over corpora and OCR — Multi-Model RAG, Invoice OCR.
Eval, safety & ops
Benchmarks, prompt defense, memory and traces — Hallucination Benchmark, Observability.
Narrative workflows: Use Cases. Additional sample walkthroughs are listed under Projects in the sidebar.
Core Capabilities
ML lifecycle (end-to-end pipeline)
Automate the full ML lifecycle locally:
- Data ingestion – Load, validate, and preprocess from files, databases, or cloud storage
- Feature engineering – Automatic creation, selection, and transformation of features
- Model training – Multi-model selection with hyperparameter tuning
- Evaluation – Metrics, visualizations, and performance comparison
- Deployment – One-click deployment with versioning and monitoring
Advanced Capabilities
Multi-step reasoning
Evaluates multiple execution approaches and selects a suitable strategy before running code
Adaptive intelligence
Automatically adapts strategy based on dataset characteristics and user feedback
Production-ready artifacts
Fully documented, tested, and versioned outputs for reliable deployment
ML Pipeline
Data Ingestion → Validation → Feature Engineering → Model Training → Evaluation → Deployment
Pipeline Features
- Auto-termination – Platform sessions auto-close after 7 days of inactivity
- Artifact preservation – Download models, reports, and logs
- Reproducibility – Track all steps, parameters, and dependencies
Data ingestion
Load data from local files, databases, or cloud storage
Validation
Schema checks, anomaly detection, and quality validation
Feature engineering
Automated creation, transformation, and selection of features
Model training
Multi-model training with hyperparameter optimization
Evaluation
Metrics, visualizations, and comparative reports
Deployment
One-click deployment to cloud, local, or edge environments
Supported ML Tasks
Neo supports a wide range of task types—from traditional supervised and unsupervised learning to modern LLM and agentic workflows. Tasks are grouped below by domain.
Traditional ML
Tabular ML
Regression, classification, clustering, time series, anomaly detection
Computer vision
Image classification, object detection, segmentation, OCR
NLP
Sentiment analysis, NER, summarization, question answering
Audio and speech
Speech recognition, audio classification, speaker identification
Modern AI
LLM fine-tuning
Instruction tuning, LoRA/QLoRA, domain adaptation
GenAI and agentic AI
RAG systems, autonomous agents, tool calling, multi-step reasoning
Advanced and multi-modal
Recommendations, transfer learning, multi-modal pipelines
Key Advantages
Speed and efficiency
- Rapid prototyping – Quick experiments and iterations
- Auto optimization – Hyperparameter tuning and caching
- Smart caching – Reuse previous computations
Quality and reliability
- Reproducibility – Track all outputs, models, and configurations
- Version control – Full history of changes
- Automated validation – Ensures pipeline correctness
Flexibility and control
- Custom constraints – Define optimization criteria
- Interactive refinement – Guide NEO with feedback
- Integration-ready – Works with existing tools and workflows
Deployment Options
Cloud platforms
AWS, GCP, Azure managed services for scalable deployment
Local development (VS Code)
Local, private, and Git-friendly development environment
On-premises
Enterprise-grade security and compliance on local infrastructure
Edge and mobile
Lightweight deployment for IoT and mobile devices
Output Formats
Docker containers
Ready-to-run containers with all dependencies included
REST APIs
Fully documented OpenAPI endpoints for integration
Pickle and ONNX
Model serialization for reuse in custom pipelines
Cloud-native
Deploy to SageMaker, Vertex AI, or other managed platforms
Feature Highlights
Autonomous execution
Self-corrects, installs dependencies, and handles errors automatically
Full transparency
View all code, decisions, and reasoning for each action
Iterative improvement
Refines models and workflows based on validation and feedback
Smart artifacts
Generates reproducible reports, metrics, plots, and models