Union.ai

Union.ai is the definitive platform for building, serving, and scaling complex, compound AI systems. It addresses the common challenge of AI models failing in production by providing a robust, scalable, and efficient orchestration layer. Built by the creators of the popular open-source orchestrator Flyte, Union.ai is designed for AI/ML engineers who need to bridge the gap between data processing and machine learning, ensure consistency across teams, and maintain full control over their development lifecycle. The platform unifies data, automates workflows, and tracks data lineage, providing a solid foundation for enterprise AI.

How to use Union.ai

The developer workflow on Union.ai is designed for speed and production readiness:

1. Define Workflows: Using the Flytekit Python SDK, developers define their tasks and workflows as pure Python code. This includes data processing, model training, evaluation, and deployment steps.
2. Local Development & Debugging: Test and debug workflows locally. Union.ai offers interactive tasks that allow for live, remote debugging in a VSCode-like environment directly in the browser, shortening development loops from hours to seconds.
3. Deploy to Union.ai: Once ready, register the workflows on the Union.ai platform. The platform's image builder can automatically build container images without needing Docker on the local machine.
4. Execute & Scale: Trigger workflows manually, on a schedule, or reactively based on events. Union.ai handles the underlying infrastructure, scaling resources up or down as needed—even to zero—to manage costs effectively.
5. Monitor & Observe: Use the Union.ai UI to monitor executions in real-time. Access detailed logs, track resource consumption (CPU, memory, GPU) per task, and observe overall cloud costs allocated by project, team, or workflow.

Core Features of Union.ai

• Optimized Performance & Scale: Achieve millisecond execution times with reusable warm containers. Scale across multiple clusters and clouds for resource availability and compliance, with massive parallelism and full-workflow caching.
• Delightful Developer Experience: Features an in-cloud image builder, self-serve secrets management, UI-based logging, interactive remote debugging, and reactive workflow automation to make development faster and more enjoyable.
• Unparalleled Efficiency: Drastically reduce costs by leveraging fractional GPUs, Spot instances with automatic recovery, and custom silicon. The platform features efficient autoscaling with Karpenter and scale-to-zero capabilities to eliminate idle resource costs.
• Enterprise-Ready Security & Governance: Deploy in your own cloud (BYOC model) so data and code never leave your VPC. It includes granular role-based access control (RBAC), high availability, and is SOC 2 Type II compliant.
• Seamless Integrations: Works with the tools developers already rely on, including major clouds (AWS, GCP, Azure), models, and frameworks like Ray, Spark, and Dask, ensuring no vendor lock-in.

Use Cases for Union.ai

Union.ai is trusted by leading companies for their most demanding AI workloads:

• Autonomous Driving: Companies like Woven by Toyota and Wayve use Union.ai to manage the immense scale of data processing, simulation, and model training required for developing autonomous driving technology, saving millions in infrastructure costs.
• Synthetic Biology: Cradle accelerates ML development for its protein design models, orchestrating complex bioinformatics pipelines with complete reproducibility and scale.
• LLM & Generative AI Operations (LLMOps): Fine-tune and serve Large Language Models (LLMs) using techniques like LoRA and QLoRA in a reproducible, resource-optimized production environment. Build and manage complex RAG (Retrieval-Augmented Generation) pipelines.
• Computer Vision: Build, train, and deploy object detection pipelines and other computer vision applications, orchestrating every step from data ingestion to model serving.
• Data & ML Platform Consolidation: Companies like Porch have migrated from tools like Airflow to consolidate their data and ML operations onto a single, unified platform for greater efficiency and consistency.

Advantages of Union.ai

• Cost Optimization: Real-time cost observability, scale-to-zero, and efficient use of spot instances drastically reduce runaway compute costs.
• Increased Developer Velocity: Features like smart caching, remote debugging, and automated image building make developers happier and more productive.
• Production-Grade Reliability: Built for mission-critical workloads with high availability and robust error handling.
• Flexibility and No Lock-In: Integrates with your existing stack and allows you to run on any cloud, avoiding vendor lock-in.
• Enhanced Security and Compliance: The "Bring Your Own Cloud" (BYOC) model ensures your sensitive data and proprietary code remain secure within your own infrastructure.

Pricing and Plans

Union.ai offers flexible pricing models to suit different needs:

• Serverless Pay As You Go: Ideal for individuals and small teams. It offers a "ship in seconds" experience without infrastructure management. New users get $30 in free credit. Pricing is based on consumption:
  • CPU Core: $0.12 / Core-Hr
  • GPU T4: $0.71 / GPU-Hr
  • GPU A100 (40GB): $4.48 / GPU-Hr
  • Memory: $0.029 / GB-Hr
• Your Cloud & On-Prem: Designed for enterprises that require data to stay within their own VPC. This plan involves a SaaS-like platform fee plus a percentage of the compute usage managed by Union.ai. Contact sales for detailed pricing.