# AI Studio

## What is AI Studio?

AI Studio is an end-to-end platform that empowers developers, researchers, and enterprises to explore, fine-tune, evaluate, and deploy large AI models—without the hassle of managing infrastructure or engineering complexity.

Whether you're experimenting with foundational models, adapting them to your domain, or scaling them for production—AI Studio provides everything in one place.

***

### Why AI Studio?

Today’s AI demands go beyond access to pre-trained models. Organizations need infrastructure flexibility, customization workflows, and robust evaluation capabilities. AI Studio addresses this with:

* **Curated Model Catalog**: Access top-performing models in text, vision, speech, and multimodal domains.
* **No-Code Fine-Tuning**: Customize models with your data using efficient techniques like LoRA.
* **Integrated Evaluation**: Benchmark functional and performance metrics before deployment.
* **Scalable Deployment**: Launch models on-demand or via dedicated endpoints with GPU-backed infrastructure.

***

### Core Capabilities

#### Model Catalog

* Browse a curated selection of foundational and fine-tuned models.
* Each model card includes:
  * Overview and intended use cases
  * Licensing and attribution
  * GitHub repository links (if available)
  * API model string and sample usage
  * Interactive playground for real-time experimentation

#### Inferencing

* Run inference directly on models using APIs or the playground.
* Control parameters such as:
  * `temperature`, `top_p`, `max_tokens`, `logit_bias`, etc.
* Transparent, token-based pay-as-you-go pricing.

#### Fine-Tuning

* Fine-tune models such as Llama-3 and Mistral using LoRA adapters.
* Upload datasets in `instruction`, `input`, `output` JSON format.
* Monitor checkpoint progress and deploy fine-tuned versions directly.
* Supports customization for specific domains such as healthcare, legal, or customer support.

#### Evaluation

* **Model Evaluation**:
  * Evaluate task performance using datasets such as MMLU, BoolQ, HellaSwag, GSM8k, and TruthfulQA.
  * Multi-language support, including Indic languages.
  * Metrics include accuracy, relevance, and ethicality.
* **Performance Evaluation**:
  * Measure latency (TTFT, inter-token, end-to-end), throughput, and token counts.
  * Configure test load parameters such as concurrency, input/output token length, etc.
  * Compare evaluation runs across models and versions.

#### Deployment

* **On-Demand Deployment**:
  * Quick, pay-per-use deployment for experimentation or low-volume tasks.
* **Dedicated Deployment**:
  * Persistent endpoints using dedicated GPUs like NVIDIA H100.
  * Recommended for high-availability production use cases.
  * Deployment cost is based on GPU time usage.
* **Management Features**:
  * Monitor deployment status
  * Bring down unused deployments to avoid unnecessary costs

***

### Who is AI Studio For?

* **Developers & Engineers**: Build and integrate AI-powered features quickly using APIs and SDKs.
* **Researchers & Data Scientists**: Experiment with model architectures and datasets without infrastructure setup.
* **Product & Business Teams**: Evaluate and compare models to inform decisions about product integration.

***

### How It Works

1. **Explore** the Model Catalog and try out models in the playground or via API.
2. **Fine-Tune** a model using your own dataset if needed.
3. **Evaluate** both task and performance metrics using built-in tools.
4. **Deploy** the best model variant with on-demand or persistent infrastructure.
5. **Monitor** usage, costs, and metrics to ensure ongoing optimization.

***

### Key Advantages

* End-to-end AI model lifecycle support in a single platform
* Access to top open-source and proprietary models
* Efficient and scalable fine-tuning using LoRA
* Built-in evaluation with task and operational metrics
* Flexible deployment and billing options
* Developer-friendly experience with interactive playgrounds and starter code

***

### Next Steps

* Get Started with Quickstart
* Browse the Model Catalog
* Fine-Tune Your First Model
* Evaluate Your Model
* Deploy to Production

***


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.cloud.olakrutrim.com/basics/ai-studio.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.