Valohai

A Valohai implementation generally follows a 5‑step onboarding and migration flow, based on Valohai’s official setup and migration documentation.

Step 1 — Prepare infrastructure & environment (Time: varies by team)

Valohai requires setting up the core components:

• A queue machine (Redis‑based job scheduler)
• Worker machines (cloud, Kubernetes, static servers, or SLURM nodes)
• Object storage (S3, Azure Blob, GCS, etc.)
  The setup depends on whether the team chooses hybrid or fully self‑hosted deployment.
  

Step 2 — Define dependencies for your ML workloads (Time: ~30 minutes)

Teams identify required packages and runtime dependencies. These can be installed during execution or pre‑baked into Docker images.

Step 3 — Create the valohai.yaml configuration (Time: ~30 minutes)

This YAML file defines how jobs run—steps, parameters, inputs, outputs, and environment requirements. It allows running existing code on Valohai with zero code changes.

Step 4 — Add parameters, metrics & tracking (Time: 1–2 hours)

Teams enhance experiments by adding configurable parameters, logging metrics, and printing results in JSON for automatic tracking.

Step 5 — Set up output artifact storage (Time: ~30 minutes)

Models and files are saved to /valohai/outputs/ and versioned automatically.

Step 6 — Update data access paths (Time: 1–2 hours)

Data locations are mapped in valohai.yaml, enabling Valohai to fetch or mount inputs automatically.

Total Estimated Implementation Time

Based on documented time references:

• Minimum: ~3–5 hours for a simple project
• Typical real‑world implementation: 1–3 days, depending on infrastructure, data pipelines, and the number of ML workflows.

Valohai is highly customizable, and most of its design philosophy is built around flexible, organization‑specific workflows. Below are the strongest data‑supported customization capabilities:

1. Full infrastructure flexibility (cloud, hybrid, self‑hosted)

Valohai can run on:

• Any cloud provider (AWS, Azure, GCP, Scaleway, Oracle)
• On‑premises data centers
• Kubernetes clusters or SLURM HPC environments
  This allows organizations to align Valohai with their infrastructure, security, or data‑residency needs.
  

2. Custom execution environments via Docker

Companies can use any Docker image, including custom-built ones with specific dependencies, GPU drivers, or frameworks.

3. Custom ML pipelines via valohai.yaml

Businesses can design complex pipelines for:

• Data ingestion
• Feature engineering
• Training
• Evaluation
• Deployment
  All customizable via YAML definitions, with no platform‑forced restrictions.
  

4. Flexible compute worker architecture

Organizations can choose:

• Autoscaled cloud VMs
• On‑prem static servers
• Kubernetes pods
• SLURM nodes for HPC
  This allows full alignment with business computing strategy.
  

5. API‑first integrations with existing systems

Valohai integrates seamlessly with:

• CI/CD workflows
• Internal automation systems
• Data platforms
• Enterprise ML tools
  All through REST APIs and CLI tools—ideal for custom internal setups.
  

6. Custom parameters, metrics & artifact handling

ML teams can define:

• Business‑specific parameters
• Custom metric logging
• Domain‑specific output artifacts (models, reports, CSVs, embeddings)
  

7. Flexible data access configuration

Businesses can point Valohai to any internal or cloud storage system and define custom data mapping in YAML.

8. Zero‑lock‑in code philosophy

Valohai does not require SDK lock‑in; it runs whatever code teams already have, in any language or framework. This is ideal for organizations with diverse legacy systems or domain‑specific tools.

Valohai does not publicly list setup fees or specific maintenance charges, but multiple official sources clarify how its pricing structure works:

1. No usage‑based or scale‑based fees

Valohai charges per user license, not based on compute usage, pipeline count, or experiment volume. This avoids hidden variable charges.

2. Support, onboarding, and training are included

All Valohai subscription plans include:

• Technical support
• Access to the Valohai Training Academy
• Onboarding resources
  

3. Infrastructure costs are separate

If the customer uses their own cloud or on‑premises setup, they bear the cost of compute (VMs, GPU instances, Kubernetes, storage).
This is not a Valohai fee but part of running ML workloads.

4. No public mention of setup or implementation fees

Neither Valohai’s pricing page nor third‑party listings mention upfront setup fees. Instead, they emphasize custom quotes, which may include enterprise-level onboarding depending on the customer.

5. Training included, but enterprise support tiers may vary

Valohai offers 24×7 support, email, phone, and knowledge base access depending on plan type, but exact pricing for higher-tier support requires a quote.

Valohai provides comprehensive training and multi‑channel support resources:

1. Valohai Academy (Formal Training Programs)

A dedicated training portal offering structured, multi‑module learning paths, including “Mastering Valohai,” hands‑on labs, and certification‑style learning.

2. Extensive documentation & how‑to guides

The official documentation includes step‑by‑step tutorials, environment setup guidance, pipeline creation, troubleshooting, and examples for computer vision, NLP, LLM fine‑tuning, and more.

3. Customer support included in subscriptions

Subscriptions include technical support from the Valohai engineering team as part of the offering.

4. 24/7 support availability (depending on plan)

Valohai provides:

• 24×7 support
• Email support
• Phone support
• Knowledge base
• FAQs/forum
  

5. Community forums

Valohai offers open forums for community discussion, Q&A, feature ideas, and collaborative support.

6. Quickstart tutorials

Official guides help users run their first experiment in minutes, including “Hello World” training and pipeline templates.

Valohai uses enterprise‑grade security frameworks, encryption, and compliance measures:

1. SOC 2 Type II Compliance

Valohai is officially SOC 2 Type II compliant, covering security, confidentiality, availability, privacy, and processing integrity.

2. Data encryption (in transit & at rest)

Sensitive data is encrypted using AWS KMS and standard cryptographic protocols.

3. Segregated production environments

Production data is segmented, and Valohai never stores source code—only temporary permission for execution.

4. Infrastructure security using AWS ecosystem

Security stack includes:

• AWS KMS
• AWS GuardDuty
• Continuous network monitoring
• Vulnerability scanning
  

5. Third‑party penetration testing

Industry‑leading security firms conduct annual penetration tests on Valohai’s codebase, deployments, and environment.

6. Strict access control features

Valohai enables:

• Organization, team, and project-level access policies
• MFA enforcement for remote access
• Least‑privilege permissions internally
  

7. GDPR compliance

Valohai is fully committed to GDPR and provides data‑processing agreements, compliance documentation, and support for customer GDPR obligations.

Valohai does not publish a fixed release schedule, but its documentation and platform architecture indicate that updates are delivered continuously, especially for customers using the Valohai‑managed cloud application.

How Updates Are Managed

1. Managed Cloud Customers
   For hybrid deployments where the Valohai application layer is hosted by Valohai (app.valohai.com), updates are applied automatically by Valohai as part of the managed service. This includes feature updates, security patches, UI improvements, and backend enhancements.
   
2. Self‑Hosted Customers
   Organizations running Valohai entirely on‑premises or in a self‑hosted environment receive updates as Docker images, which they can deploy on their own schedule. This gives enterprises full control over update timing, especially for regulated or air‑gapped environments.
   
3. Documentation Updates
   Valohai frequently updates its documentation and publishes platform improvements or changes, signaling an ongoing cycle of product enhancement.

Valohai has a clear stance on data ownership: customers retain full control over their data, including datasets, models, experiment outputs, and metadata stored in their infrastructure.

Key Data Ownership Principles

1. Customer‑Owned Data (Controller Model)
   Valohai states that its customers act as data controllers, defining how data is processed and maintained. Valohai acts solely as a processor, assisting in data handling but holding no ownership rights.
   
2. No Direct Access to Customer Data
   Valohai does not store or retain customer datasets; all data stays within the user's environment (cloud or on‑prem). Valohai only temporarily accesses items required for job execution and never stores source code.
   
3. Data Portability
   Because Valohai uses open standards (Docker, YAML, S3‑compatible storage), all assets—models, logs, datasets, metadata—remain in portable, standard formats. Customers can export, migrate, or offboard without vendor lock‑in.
   
4. GDPR Compliance
   Valohai’s GDPR documentation reinforces that the customer controls:
   • Purpose of data processing
   • Retention
   • Deletion
   • Transparency to data subjects.

Valohai is intentionally designed to scale flexibly with evolving organizational needs, both technically and commercially.

1. Commercial Scaling (Licensing Model)

Valohai pricing is based per user license, not compute usage or workflow volume. This means organizations can scale up or down by simply adjusting user seats.

2. Technical Scaling (Compute Resources)

Valohai automatically supports scaling through several options:

• Auto‑scaling compute resources (cloud VMs, Kubernetes pods, GPU clusters) that adjust based on workload demand.
• Support for multiple infrastructures, including VMs, Kubernetes, SLURM clusters, and static servers—making it easy to scale horizontally or vertically as needs grow.
  

3. Scaling Down

Because compute is separate from Valohai licensing, organizations can scale down by:

• Reducing or removing compute workers
• Lowering the number of active user accounts
• Pausing workloads or retiring pipelines
  There are no scale‑related platform penalties since pipelines, deployments, and experiments are unlimited.
  

4. Infrastructure‑Agnostic Scaling

Customers can switch from:

• Cloud → On‑prem,
• On‑prem → Cloud,
• Hybrid → Fully self‑hosted,

without architectural lock‑in.

Valohai meets several enterprise-grade compliance and security standards, supported by official documentation:

1. SOC 2 Type II Compliance

Valohai is officially SOC 2 Type II certified, covering all 5 Trust Service Principles:

• Security
• Availability
• Processing integrity
• Confidentiality
• Privacy
  

SOC 2 Type II certification requires continuous monitoring and long-term adherence to audited security practices.

2. GDPR Compliance

Valohai is fully aligned with the EU GDPR and provides:

• GDPR‑compliant terms
• Privacy policies
• Data Processing Agreements
• Controller/Processor role definitions
  

Valohai emphasizes helping customers meet GDPR obligations through documentation, training, and system design.

3. Enterprise‑grade security controls

Valohai implements strong data protection measures, including:

• Encryption at rest and in transit (AWS KMS)
• Segmented production environments
• Continuous monitoring & vulnerability scanning
• Least‑privilege access
  

These are consistent with SOC2‑aligned enterprise requirements.

4. Third‑party penetration testing

Valohai undergoes regular professional penetration tests covering:

• Application security
• Source code review
• Infrastructure penetration testing
  

5. MFA enforcement & secure access

Valohai enforces MFA, encrypted communication channels, annual security training, and strict internal access control.