OxVillage Logo

    Tensorrr

    A verification platform that combines AI reasoning with actual code execution to factually validate technical quality.

    Focus/AI verification and technical due diligence

    Context/Built from VU DLAB

    Visit Website
    Tensorrr website screenshot

    The Challenge

    AI-generated code and inflated resumes make it nearly impossible for recruiters and investors to distinguish who can actually build from who just sounds smart. Existing code analysis tools lack context, hallucinate features, and can't verify whether something truly works — or only exists on paper.

    How it works

    Tensorrr runs and tests GitHub repositories in isolated sandbox environments. Where standard tools only read code, Tensorrr actually executes it — and validates AI findings with deterministic checks.

    Code execution

    Repositories are built and run inside a sandbox. Claims aren't assumed but proven by actually executing the code.

    Context-aware analysis

    The system maintains project-wide context: it understands architecture patterns, dependency relationships, and how modules interact.

    Multi-layer verification

    AI reasoning is cross-validated against code output. Hallucinations are eliminated because every finding must have a factual basis.

    Real-time insight

    During analysis, the full reasoning process is streamed live, so users watch the audit unfold in real time.

    Under the hood

    Frontend

    A responsive Next.js interface with real-time streaming of the analysis process via WebSocket and 3D visualizations using WebGL.

    Intelligence layer

    Multi-agent AI workflows via the Claude API, with isolated sandbox execution through E2B. Every AI finding is verified by actual code runs.

    Backend

    Supabase with Row-Level Security, a Node.js orchestration layer for agent coordination, and real-time communication via Socket.io.

    Tensorrr is live in production and actively used for recruitment assessments.

    Traction

    In production

    Fully functional audit system, live and available

    Proven technology

    Complex AI workflows with real-time verification, successfully deployed

    Market ready

    Actively used for technical assessments in recruitment

    Scalable

    Architecture built for enterprise-grade due diligence

    A concrete example of this kind of audit report can be viewed on my LinkedIn.

    What this demonstrates

    • Design and implementation of multi-agent AI workflows with factual verification
    • End-to-end ownership of a SaaS platform — from real-time UI to distributed AI orchestration
    • Production-grade systems that deliver direct value to real users

    Built with a modern stack

    Frontend

    • Next.js
    • React
    • TypeScript
    • Tailwind CSS
    • OGL (WebGL)

    Backend

    • Supabase (PostgreSQL, Auth, RLS)
    • Node.js
    • Deno (Edge Functions)

    Realtime / Communication

    • Socket.IO
    • Supabase Realtime

    AI

    • Claude API, Claude Agent SDKs
    • Prompt Engineering
    • Tool / Function Calling
    • Agent Orchestration
    • E2B (sandboxed code execution)

    Tools / DevOps

    • pnpm (Monorepo)
    • Docker (E2B template)
    • CI/CD (deployments via Vercel & Railway)