Legal Documentation

Terms of Service

Please read these terms carefully before engaging with our AI project recovery, audit, optimization, or consulting services.

Version 1.2Last updated: February 2025

01Introduction

These Terms of Service (“Terms”) govern your access to and use of all services provided by OptyxStack (“we”, “us”, “our”).

By engaging with our services, signing a proposal, or accessing any part of our platform, you agree to these Terms. If you do not agree, you must not use our services.

02Scope of Services

OptyxStack provides AI project recovery and optimization services, including but not limited to:

  • AI System Audit (diagnostic, baseline, failure analysis)
  • AI Recovery Sprint (fix accuracy, hallucinations, production failures)
  • AI Optimization & Reliability (cost, latency, drift detection)
  • RAG Optimization (retrieval, reranking, groundedness)
  • LLM Evaluation & Golden Set Development
  • Before/After Benchmarking & ROI Measurement
  • Prompt, Routing, Caching & Context Optimization
  • Regression Gates & Monitoring Dashboards
  • Vendor-Built AI System Second-Opinion Audits

Service details may vary depending on the chosen engagement, contract, or proposal.

03Engagement Requirements

To perform our work effectively, the client agrees to:

  • Provide timely access to staging environments, logs, metrics, or sample conversations when required for audit or recovery work.
  • Assign at least one point of contact for communication.
  • Ensure that systems, applications, and data provided for work are legally owned or licensed.
  • Provide read-only access to logs and telemetry where applicable; we can operate on redacted or anonymized data where needed.

We are not liable for delays caused by missing or late access/information from the client side.

04Data & Access

4.1. Access Requirements

We typically require staging access, read-only logs/metrics, and sample conversations (or redacted equivalents) to perform audits and recovery work. We do not require production write access or customer data unless explicitly agreed.

4.2. Data Use & Training

No customer data is used to train models. We operate on staging, redacted datasets, or sample conversations. Client data is used solely for the purpose of the engagement and is not retained beyond the agreed handoff period unless otherwise specified.

4.3. Least-Privilege

We work with least-privilege access where possible and can operate under NDA on redacted or anonymized data to meet your security and compliance requirements.

05Payment & Billing

Payment Terms

  • • Fixed-scope audit: typically paid before or at engagement start.
  • • Recovery sprint: priced after baseline; billed per milestone or monthly.
  • • Retainer: billed monthly in advance.
  • • Payments must be completed as per the signed proposal or contract.

Refund Policy

  • • AI Audit: Non-refundable once work has begun.
  • • Recovery Sprint: Non-refundable for completed milestones.
  • • Retainer: prorated refund only for unused prepaid time, at our discretion.
  • • Refunds only for clear billing errors.
Late payments may result in service suspension or pause of deliverables after a warning period.

06Project Scope & Changes

Work is carried out based on the agreed scope. Any new requests (additional recovery goals, model changes, expanded eval sets, or scope beyond the baseline) may require additional fees, a new timeframe, or an updated proposal. We will not expand scope without client approval.

07. Client Responsibilities

  • Maintaining valid access credentials (staging, logs, dashboards).
  • Ensuring data provided is legal, compliant, and properly licensed.
  • Informing us before major prompt, model, or pipeline changes during an active engagement.

08. Confidentiality

Both parties agree to keep all business, system, and customer data confidential.

Separate NDA available upon request.

09Security & Data Protection

OptyxStack follows industry-standard security practices. However:

  • 1.No system can guarantee absolute security.
  • 2.Client data is used solely for the engagement; we do not use it to train models.
  • 3.The client is responsible for application-level vulnerabilities unless covered by a separate security engagement.
  • 4.We are not liable for breaches caused by third-party LLM providers, APIs, or client-side misconfigurations.

10Intellectual Property

Code provided by the client:Remains property of the client.
Code, evals, or tools created by OptyxStack:We grant the client a license to use any optimizations, eval harnesses, golden sets, scripts, or dashboards created for the project. Internal tools, libraries, or automation frameworks built by us remain exclusively ours unless stated otherwise in the proposal.

11Suspension & Termination

We may suspend or terminate services if payments are overdue, the client violates laws, or access is misused. The client may terminate with notice according to the contract. Refunds are not provided for completed audit work or delivered milestones.

12Limitation of Liability

To the maximum extent permitted by law:

  • OptyxStack is not liable for indirect damages, revenue loss, downtime, or business disruption.
  • Total liability is limited to the amount the client has paid in the past 30 days.
  • We are not responsible for issues caused by third-party providers (LLM APIs, cloud providers, vector DBs, etc.).

13. SLAs (If Applicable)

SLAs regarding response times, deliverables, or milestones only apply if explicitly included in the service contract or proposal.

14. Communication

We use Email and agreed channels (Slack, etc.). Weekly readouts and support cadence depend on your engagement type.

15. Modifications

We reserve the right to update these Terms. Continued use of services indicates acceptance.

16. Governing Law

These Terms are governed by the laws of the jurisdiction in which OptyxStack operates unless stated otherwise.

17. Contact Us

Support Email