Roadmap
Planned features and development priorities for AiDotNet.
Current Focus
Q1 2026
Documentation & Developer Experience
- Comprehensive tutorials for all feature categories
- 30+ sample applications
- End-to-end application examples
- GitHub Pages documentation site
- Interactive API documentation
- Video tutorials
Performance
- SIMD optimizations for tensor operations
- Memory
migration for zero-copy - Additional GPU kernel optimizations
- Intel MKL integration
- ARM NEON optimizations
Short-Term (Q2 2026)
HuggingFace Integration
- Direct model loading from HuggingFace Hub
- Tokenizer support for popular models
- Model card parsing
- Automatic weight conversion
Training Improvements
- Mixed precision training (FP16/BF16)
- Gradient accumulation optimizations
- Memory-efficient attention
- Flash Attention support
New Models
- LLaMA 3 support
- Mistral/Mixtral support
- Phi-3 models
- Claude tokenizer compatibility
Medium-Term (Q3-Q4 2026)
Distributed Training
- Multi-node training improvements
- Elastic training (fault tolerance)
- Improved checkpointing
- Cloud provider integrations (Azure ML, AWS SageMaker)
Computer Vision
- YOLO v12 (when released)
- Grounding DINO
- Open vocabulary detection
- Video understanding models
Audio
- Whisper large-v3 turbo
- Real-time transcription
- Speaker diarization
- Music understanding models
LLM Features
- Speculative decoding
- KV cache optimizations
- Continuous batching
- PagedAttention
Long-Term (2027+)
Multi-Modal
- Vision-Language models
- Audio-Language models
- Multi-modal embeddings
- Multi-modal RAG
Edge Deployment
- WASM support
- Mobile optimization (iOS, Android)
- IoT deployment
- TinyML support
Research Features
- Neural Architecture Search improvements
- AutoML v2 with more algorithms
- Federated learning
- Privacy-preserving ML
Ecosystem
- Visual Studio extension
- Jupyter kernel
- MLOps integrations
- Model registry
Feature Requests
Want to see a feature on the roadmap? Here’s how:
- Check existing issues: Search GitHub Issues first
- Open a feature request: Use the feature request template
- Discuss in GitHub Discussions: For broader ideas
- Vote on existing requests: Add a thumbs up to prioritize
How Priorities Are Set
Features are prioritized based on:
- Community demand - Number of requests and votes
- Impact - How many users benefit
- Alignment - Fits AiDotNet’s mission
- Feasibility - Technical complexity
- Contributor interest - Available resources
Contributing to the Roadmap
Want to help implement a roadmap item?
- Comment on the related GitHub issue
- Discuss your approach
- Submit a PR when ready
See the Contributing Guide for details.
Completed Milestones
2025
- Initial release with 100+ neural networks
- GPU acceleration via CUDA and OpenCL
- 106+ classical ML algorithms
- Computer vision models (YOLO, DETR)
- Audio processing (Whisper, TTS)
- RAG components
- LoRA fine-tuning
- Distributed training (DDP, FSDP, ZeRO)
Early 2026
- Comprehensive documentation site
- 30+ sample applications
- End-to-end examples
- Performance optimizations (SIMD, Memory
)