About

    Umer Farooq builds the production layer around AI agents.

    Zenovae is Umer Farooq's studio. We embed with technical teams, ship agents into their repo, and score whether those agents do the right thing. We are based in Islamabad and work with US teams. The overlap hours are confirmed on the scoping call.

    Start a scoping call

    3+

    Years AI-Focused

    40+

    Tool Classes

    7-14d

    First Build

    24/7

    Ops Posture

    Operating Principles

    How we decide what to build and how to build it.

    These principles shape every architecture decision, every integration surface, and every handoff.

    Production beats demo polish.

    Every system we ship is designed to hold up under real load, real users, and real edge cases. We do not optimize for the screenshot.

    Every tool action gets a boundary.

    We define permissions, retries, approval gates, and rollback paths before the first API call. Safety is not an afterthought.

    Retrieval is evaluated, not assumed.

    RAG pipelines get benchmarked against real question sets with measurable accuracy. We know the precision, not just the architecture.

    Costs and latency are design constraints.

    We choose models, chunking strategies, and caching layers with budgets in mind. Performance is part of the spec.

    Humans stay in the loop where risk demands it.

    The system knows when to ask, not just when to act. We design escalation paths, not black boxes.

    Expertise

    Tech stack.

    The tools we use to build systems that ship.

    LLM Platforms

    OpenAI GPT-4
    Anthropic Claude
    Google Gemini
    Together AI
    Mistral

    Frameworks

    LangChain
    LlamaIndex
    LangGraph
    Semantic Kernel
    Haystack

    Vector DBs

    Pinecone
    Weaviate
    Qdrant
    ChromaDB
    pgvector

    Backend

    Python
    FastAPI
    Node.js
    TypeScript
    Go

    Frontend

    React
    Next.js
    TypeScript
    Tailwind CSS
    Vite

    Cloud

    AWS
    GCP
    Vercel
    Docker
    Kubernetes

    Databases

    PostgreSQL
    MongoDB
    Redis
    Elasticsearch
    Supabase

    ML/AI

    PyTorch
    Hugging Face
    scikit-learn
    TensorFlow
    ONNX

    Start a scoping call.

    Tell us whether you need an audit, a build, or an embedded engineer.