Skip to main content

TrainLab AI Documentation

Welcome to the TrainLab AI documentation portal. TrainLab AI is a comprehensive no-code platform that lets anyone turn raw data into production-grade AI models—without MLOps heavy lifting. Upload or connect data, validate and label, fine-tune the right open-source model, and deploy an API/chatbot in minutes—with costs and time auto-estimated up front.

🚀 Platform Overview​

TrainLab AI provides an intuitive drag-and-drop interface for creating, training, and deploying machine learning models. Our complete workflow takes you from raw data to deployed APIs through a streamlined 10-step process.

Key Features​

  • No-Code Interface: Build AI models through an intuitive visual interface
  • Multi-Industry Support: Pre-configured templates for 20+ industries
  • Flexible Data Formats: Support for CSV, JSONL, and image formats
  • Smart Training Plans: Auto-tuned hyperparameters based on your data and domain
  • Transparent Pricing: Live cost and time estimates before you commit
  • Enterprise Ready: SOC 2 Type II, GDPR, and industry-specific compliance
  • Scalable Infrastructure: From proof-of-concept to enterprise deployment

📋 How TrainLab AI Works (10 Steps)​

1. Connect or Upload Data​

Plug in S3/Redshift/MySQL/Kaggle/HF or upload CSV/JSONL/ZIP. Run automatic schema detection, PII flags, and quality checks.

2. Validate & Encrypt​

Run warnings/errors, fix missing fields, and optionally encrypt data at rest (per-Datasource policy) before use.

3. (Optional) Label & Augment​

Use the built-in Annotation Studio (text, vision, multimodal) with roles, consensus rules, and progress KPIs.

4. Pick a Blueprint​

Choose from template tasks (LLM SFT/DPO/ORPO, extractive QA, VQA, captioning, image classification, tabular regression/classification, ST pair/triplet, etc.). Blueprints pre-fill model/task defaults.

5. Model Compatibility Check​

We auto-filter models by token/window size, modality, and license. You can bring your own HF model too.

6. Smart Training Plan​

Our parameter optimizer proposes batch size, LR schedule, epochs, LoRA/PEFT, grad checkpointing—tuned to dataset size, domain, and GPU memory.

7. Transparent Cost & Time​

Get a live estimate for runtime and spend (dataset size × model × GPU). Adjust sliders; see impact instantly.

8. Train with Guardrails​

Launch fine-tuning on our managed GPUs. Real-time logs, eval metrics, early-stop, resume, and checkpoint policies included.

9. Evaluate & Approve​

Compare runs, browse artifacts, review confusion matrices/ROUGE/BLEU/MAP, and promote the best checkpoint.

10. Deploy & Scale​

One-click deploy to vLLM/Triton inference. Get REST/Chat APIs, keys, rate limiting, observability, and usage analytics. Share privately—or list on the Marketplace.

🔧 Platform Components​

Datasources​

  • Connect anything: S3, Redshift, MySQL/Postgres, Hugging Face, Kaggle, or upload CSV/JSONL/ZIP
  • Quality & compliance: Schema mapping, type checks, duplicate detection, PII tagging & optional encryption
  • Versioned & auditable: Raw vs processed lineage, who changed what, and when

Annotation Studio​

  • 23+ task types: Text, image, multimodal (VQA, captioning), ST pair/triplet
  • Quality controls: Consensus, gold sets, inter-annotator agreement
  • Operations: Role-based queues, shortcuts, keyboard nav, and progress KPIs

Blueprints (Templates)​

  • Pre-wired setups: SFT, DPO/ORPO, extractive QA, image classification, tabular classification/regression
  • Opinionated defaults: Data splits, augmentations, loss/metrics tuned per task
  • Bring your own: Start from any HF model or your saved base

Training & Deployment​

  • Infrastructure: Managed GPUs
  • Live telemetry: Loss curves, grad norm, LR, eval metrics; step/epoch logs
  • Reliability: Resume from checkpoints, early stop, save-total-limit, artifact retention
  • Backends: vLLM/Triton with quantization options

Marketplace & Playground​

  • Try it live: Prompt, upload, or demo sets; share links
  • Embed anywhere: JS widget, no-code web components
  • List models: Publish with sample IO, docs, and pricing
  • Discover: Filter by task, domain, latency, and rating

📚 Documentation Structure​

Our documentation is organized by AI Tasks and Industries to help you quickly find relevant information for your use case.

Browse by Task​

Language Models (LLM)​

Vision Models​

Computer Vision​

Text Analysis​

Document Processing​

Structured Data​

Browse by Industry​

🎯 Quick Start Guide​

1. Choose Your Task​

Select the AI task that matches your business needs from our comprehensive task library.

2. Select Your Industry​

Pick your industry to access pre-configured templates and compliance settings.

3. Connect Your Data​

Use our Datasources to connect S3/databases or upload files directly.

4. Validate & Prepare​

Run quality checks, handle PII, and optionally use Annotation Studio.

5. Configure Training​

Pick a Blueprint, review compatibility, and approve the training plan.

6. Train Your Model​

Launch fine-tuning with real-time monitoring and cost tracking.

7. Deploy and Monitor​

One-click deploy to production APIs with observability and scaling.

📊 Supported Data Formats​

Text Data​

  • CSV: Comma-separated values with headers
  • JSONL: JSON Lines format for complex structured data

Image Data​

  • Supported Formats: JPG, JPEG, PNG
  • Delivery Method: ZIP file containing images + CSV/JSONL mapping file

Database Connections​

  • Supported: S3, Redshift, MySQL, PostgreSQL, MongoDB
  • External: Hugging Face, Kaggle, GitHub repositories

👥 Who TrainLab AI is For​

Data/Domain Teams​

Upload, clean, and label data; approve quality and compliance.

ML Engineers​

Tune parameters, compare runs, and ship best checkpoints to production.

Developers​

Consume APIs/SDKs, embed chat widgets, and wire automations.

Leads/Compliance​

Review costs, timelines, and audit trails before approvals.

🔒 Security & Compliance​

TrainLab AI maintains industry-leading security standards and compliance certifications:

Trust & Security​

  • Data Control: Your buckets or ours; encryption optional per Datasource
  • Compliance Ready: Clear lineage, role-based access, and signed deployments
  • Observability: Request logs, redaction options, and per-key analytics

Certifications​

  • SOC 2 Type II
  • GDPR Compliant
  • HIPAA Compliant (Healthcare)
  • PCI DSS (Finance)
  • ISO 27001

📞 Support​

🚦 Getting Started​

Ready to begin? Start your AI journey today:

TrainLab AI - Empowering industries with Artificial Intelligence