Foundations · grounding & infrastructure

Grounded in real facts.
Built with zero Sutherland data.

Two foundations carry this prototype. First, every load-bearing fact is grounded in public, verifiable sources — no speculation, no hallucination. Second, the whole thing runs on public cloud infrastructure with no connection to any Sutherland system, dataset, or real candidate.

Adversarially fact-checked against primary sourcesCloudflare free tierSynthetic data only
The data boundary

Public in. Synthetic inside. One real call out.

Everything inside the dashed boundary runs on uberstudio.io. The only data that ever enters is public documentation and synthetic personas. The only real external call that leaves is to Anthropic Claude. Nothing from Sutherland crosses the line.

No Sutherland dataSynthetic personas1 real external dependency
Explore the two pillars

What it stands on — pick a pillar, open any card.

Pillar A is the resources used to ground the work in real-world facts and the latest public AI capabilities. Pillar B is the infrastructure used to build the prototype without any Sutherland data.

The simulated ATS mirrors the published contracts so the prototype could flip to live by swapping a binding, with no logic rewrite.

  • OAuth authorization the real platform uses
  • Status enum: LEAD → NEW → IN_REVIEW → INTERVIEW → OFFERED → …
  • Status-history timestamps — “cold” is COMPUTED from time-in-stage; there is no native cold event
  • Internal notes API for writing recruiter-visible context back

Where the source deck mis-stated figures, the build replaced them with the vendor’s own published, verifiable numbers.

  • 2× conversion with Winston Chat (published)
  • 26% faster hiring (published)
  • 75% less time screening (published)
  • Winston remains the in-chat assistant persona

The simulator enforces the same mechanics as the live platform so compliance and UX behave identically.

  • Template messages to re-open a conversation
  • The 24-hour customer-service window for free-form replies
  • Interactive Flows for in-chat scheduling and capture
  • Opt-in requirements and per-message pricing model

Every statistic that justifies the build is traceable to a published source — used to scope the problem, not to embellish it.

  • ~60% start an application but never finish
  • 61% ghosted after an interview
  • 57% abandon the process mid-way
  • No number appears without a verifiable source
The discipline

Zero speculation. Every load-bearing fact verified.

Where a source — including the vendor whitepaper — stated a number that could not be substantiated, it was corrected to the published, verifiable figure or dropped. The result is a prototype whose claims you can trace back to primary sources, not to invention.

No Sutherland systems

Nothing connects to internal Sutherland infrastructure.

No Sutherland data

No proprietary datasets, exports, or configs are used.

No real candidate PII

Every persona in the demo is fully synthetic.

Public docs only

Behaviour is modelled from publicly published contracts.

Keep exploring

See the workflow, or try the live prototype.

Head back to the executive walkthrough, open the recruiter dashboard, or return home.

← Under the HoodOpen the dashboard →Home