# FinetuneDB

> AI fine-tuning platform to create custom LLMs by training models with your data in minutes, not weeks.

FinetuneDB is an AI fine-tuning platform that enables teams to create custom large language models (LLMs) by training them on proprietary data. The platform streamlines the entire workflow from dataset building to fine-tuning, deployment, and ongoing evaluation, all from one secure interface. It helps organizations achieve better model performance at lower cost by customizing AI models for specific use cases.

- **Dataset Management** allows teams to collaboratively build proprietary fine-tuning datasets with a collaborative editor, enabling multiple team members to work together on creating high-quality training data that differentiates model performance from competitors.

- **Model Evaluation** provides human and AI feedback capabilities including automated Copilot evaluations, benchmarking with domain experts and engineers, and metrics tracking for speed, quality scores, and token usage.

- **Production Logging** captures user interactions, model responses, and system metrics for ongoing model improvement, with advanced filters for precise searching and tracing for visualizing language chains with unlimited nested actions.

- **Prompt Studio** offers a prompt playground for creating, managing, and optimizing prompts, with version tracking and the ability to compare foundation models against fine-tuned versions.

- **Developer Integration** includes SDKs for Python and JavaScript/TypeScript that extend the OpenAI SDK, allowing quick integration within minutes. The platform also supports Web API for custom integrations and Langchain callback handlers for full execution traces.

- **Serverless Inference** provides pay-as-you-go pricing for models including Llama 3, Llama 3.1, Llama 3.2, and Mixtral variants with competitive inference and fine-tuning pricing.

- **Security Features** include TLS 1.2+ encryption in transit with perfect forward secrecy, AES 256 encryption at rest, strict role-based permissions management, and SOC 2 compliance in progress. All data is stored on European servers.

To get started, sign up for a free account to access one project and dataset with 1,000 logs per month and $10 in free credits. Upload existing datasets in JSONL format or integrate the SDK to capture production data. The platform supports bringing your own API keys from providers like OpenAI and Anthropic, and allows integration with self-hosted models via vLLM.

## Features
- Collaborative dataset building
- Model evaluation with human and AI feedback
- Production data logging and tracing
- Prompt playground and version management
- OpenAI SDK compatible integration
- Python and JavaScript/TypeScript SDKs
- Langchain integration
- Serverless inference
- Advanced filtering and search
- Custom evaluation workflows
- Role-based access control
- Data encryption in transit and at rest
- JSONL dataset format support
- vLLM integration for self-hosted models

## Integrations
OpenAI, Anthropic, Langchain, vLLM, Llama 3, Llama 3.1, Llama 3.2, Mixtral

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://finetunedb.com
- Documentation: https://docs.finetunedb.com/introduction
- EveryDev.ai: https://www.everydev.ai/tools/finetunedb