ReflexAxis
For Company Owners

We train AI models with multipurpose datasets.

Our RL environments help push the boundaries of frontier AI.

We work across sectors such as manufacturing, engineering and construction.

Abstract illustration of glass data modules arranged on a circuit-style grid.
Our work

AI Environments and Evals

We build RL environments for frontier AI agents. In these environments, models complete operational tasks to improve business outputs. Our diverse and high-fidelity training gyms accurately model the real business world and all of its work for AI.

Job and work records

Work orders, job notes, quotes and change orders.

Email and chats

Threads with customers, vendors and internal team.

Documents and files

Proposals, checklists, procedures and drawings.

Accounting and billing

Invoices and estimates for vendors and customers.

Modeled on the most commonly used tools
  • Salesforce
  • HubSpot
  • Slack
  • Outlook
  • Microsoft Teams
  • Gmail
  • Google Drive
  • QuickBooks
  • Xero
  • Dropbox
  • Excel files
  • Linear
  • GitHub
  • And most others

thesis

Pushing the frontier forward

We're working on what it takes to build AIs that can assist in meaningfully moving the needle on real business outputs, not token usage for the sake of vanity metrics.

  • We're working to improve model benchmarks
  • We develop evals that improve output
  • Agents iterate on outputs and failures