TReqs provides an AI process platform that automatically records experiment lineage, inputs, outputs, and environment metadata via its roar CLI and a hosted GLaaS registry. The system offers searchable, live lineage visualizations, cost attribution, and policy enforcement tools to help machine learning teams reduce compute waste, streamline audits, and collaborate securely.
Funding
Funding not disclosed
Founders
Product
Problem
Machine learning teams often struggle to keep track of experiment lineage, decisions, data, and resource usage, leading to wasted runs, difficult audits, and compliance challenges.
Solution
TReqs offers an AI process platform that automatically captures lineage through its roar CLI and hosted GLaaS registry, providing a continuous, queryable record of experiments. The platform implements the TRAC framework—ensuring that all decisions, changes, and data are traceable, reproducible, attributable, and collaborative. By centralizing lineage and enabling asynchronous collaboration, TReqs reduces cycle time, cuts unnecessary compute spend, and simplifies audit and compliance workflows. Integrated features such as live lineage, cost and compute attribution, policy approvals, and Slack/webhook notifications give teams clear governance without slowing development. The solution scales from individual free users to enterprise teams with role‑based access, SSO, and dedicated support.
Target Audience
Primary customers are machine learning engineering teams in startups, mid‑size companies, and enterprises that need reliable provenance, cost tracking, and collaborative workflow management.
Features
- roar command‑line tool that automatically records runs, inputs, outputs, and environment metadata
- GLaaS (Git‑like AI lineage as a service) registry with public or private lineage storage and unlimited retention
- Live lineage visualization and searchable history across projects and nodes
- Cost and compute attribution linking spend to specific experiments and resources
- Policy engine with approvals, customizable rules, and Slack/webhook integration for governance
- Multi‑node capture (Ray) and cloud integrations (AWS S3, Weights & Biases) for distributed training
- Role‑based access control, SSO/SAML, and audit logs for enterprise security
- Free reader seats for stakeholders to view results and lineage without write permissions