CloudFunnelCloudFunnel
CloudFunnelApplied AI R&D · Part of Team Drsti

The R&D lab building Drsti.

CloudFunnel is the applied-research and engineering arm of Team Drsti. We design and build Drsti — governed, multi-tenant AI — and pressure-test how AI should actually build software. We run the hard parts on ourselves first.

How we work

One AI writes the code. A different AI reviews it.

Everything we build runs on the same discipline: explicit intent before code, gated steps, and a cross-model review where the model that checks the work is not the model that wrote it. A model reviewing its own output has the same blind spots that produced it. That's how AI-generated code earns its way into a security-critical product — and it's the thread that runs through everything below.

What we build & research

Our research and our product are the same work. We build Drsti in the open with these three threads — and what we learn building it is the lab's real output.

What we build

Governed, multi-tenant AI

Drsti Studio is our primary build — governed, tenant-isolated AI execution with policy controls, audit trails, and human approval gates. It is where our research meets production.

  • Tenant isolation by design
  • Policy controls & audit trails
  • Human-in-the-loop approvals

How we build it

Cross-model adversarial review

One model writes the code; a different model reviews it cold against the spec with a hostile checklist. It is the discipline that lets AI-generated code earn its way into a security-critical product.

  • Implementer and reviewer differ
  • Hunts security & tenant-isolation bugs
  • Caught before merge, not after

What we research

AI-augmented SDLC

We pressure-test how AI should actually build software — gated spec → plan → implement, tiered rigor, and the operating model that keeps speed from costing correctness. We run it on ourselves first.

  • Gated spec → plan → implement
  • Tiered rigor by risk
  • Proven on our own codebase

We build it before we believe it

We don't research AI from slides. The operating model and the review discipline we rely on are the ones we use to build real, multi-tenant, security-critical software today — with AI doing much of the implementation under a disciplined review structure. When something doesn't survive contact with production, it doesn't ship.

Part of Team Drsti

CloudFunnel is the R&D and engineering arm of Team Drsti — the unified brand that also includes Nika Partners (talent) and Knowillence (delivery). Together we research, build, and stand behind Drsti.