Garnet
The engineer behind your private AI

The engineer behind your private AI.

Garnet — the AI that runs on your own computer, where your words never leave your machine — is built and operated by one engineer. The person who designed it is the person who answers when you need it set up.

Watch it build ▸ LinkedIn ↗
NYC 15+ yrs on-device AI
Live model orchestration

Watch it build.

Pick something to build as a private, on-device AI. One prompt fans out across the model stack — a reasoning model writes the plan and an image model paints it, live, while a voice reads it aloud and a 3D crystal turns alongside. Generated once, then served from cache, so it stays fast and costs next to nothing to run.

Why it's built this way

One engineer. Your machine. No middleman.

"The whole promise of Garnet is that your words never leave your computer. The way you keep a promise like that is to build the thing yourself, end to end, and answer for it personally."

Garnet is a private AI you own — a real assistant that runs entirely on your own computer, keeps working with the internet unplugged, and keeps your words on your machine. No account, no subscription, no company on the other end keeping a copy.

It's designed, built, and supported by one engineer. The person who chose the architecture is the person who installs it on your computer and answers when you have a question. Fewer hands between the promise and the product means the promise is one someone is actually accountable for.

Engineering surface area

Six disciplines. One operator.

The engineering breadth behind a product that has to run anywhere, on your own hardware, with nothing leaking out. Each was shipped on real production systems before it shaped Garnet.

💎

D365 F&O Architecture

X++, SysOperation batches, extension-only patterns. Fortune-500 supply-chain rollouts processing 500K+ daily transactions, zero overlayering risk.

☁️

Cloud Infrastructure

Azure, AWS, hybrid cloud at scale. Distributed systems with measurable SLOs, not aspirational uptime claims.

🤖

On-device AI

Local LLM inference, RAG, and quantized models tuned to run fast on everyday hardware — the engineering that makes a private AI feel instant on your own machine, with nothing sent to a server.

📊

Data Engineering

ETL pipelines, real-time analytics, data architecture. Power BI dashboards and Azure SQL pipelines for data-driven growth, not vanity dashboards.

🔐

Security & Compliance

Enterprise-grade security, audit trails, regulatory compliance. Real attestable evidence trails — not theatre.

🎨

Full-Stack Development

Backend APIs to premium frontend experiences. WebGL, three.js, the works — including the live demo on this page, built and operated in-house.

The practice

What production-grade systems require.

  1. 01

    Short feedback loops

    The engineer who answers your Discord ping is the engineer who wrote the code and ships the fix. No handoff queue.

  2. 02

    End-to-end ownership

    Decisions, code, deploys, and incident response stay inside the same engineering process — fewer dropped contexts between layers.

  3. 03

    Measurable quality

    Every lane ships with logs, metrics, and runbooks. If it can't be measured, it doesn't go to production.

  4. 04

    Tight iteration cycles

    Small batch sizes, frequent deploys, reversible changes. Production moves in days, not quarters.

  5. 05

    Coherent system design

    The on-device model, the setup flow, the website, the demo you just watched — engineered to work together. All measurable, all owned, all accountable to one person.

Origin & evolution

From hardware hacking to enterprise orchestration.

Fifteen years. One trajectory. Real ledger of ships, scales, and learnings.

2016 The computer surgeon

Building software and hardware with surgical precision. Maintaining a fleet of Ethereum miners — understanding the physical layer of computation before mastering the logical.

Hardware · GPU optimization · Mining ops
2017 The entrepreneur

Co-built ThinkJuice, a digital marketing agency, alongside Troy Osinoff and Michael Lisovetsky. Learned how to scale business value through technology, and how to scope down before scoping up.

Digital agency · ThinkJuice · Scale
2018 System genesis

Distributed systems. High-availability web applications. Production-grade engineering work — the kind that has runbooks, alerts, and postmortems — begins.

Distributed systems · High availability
2023 Business intelligence

Architected Power BI dashboards and Azure SQL ETL pipelines at Allstar Innovations. Dashboards that drove decisions, not dashboards that drove dashboards.

Power BI · Azure SQL · Analytics

Building Garnet — a private AI you own that runs entirely on your own computer. The whole product, from the on-device model to the optional done-for-you setup, is built and operated by one engineer, so the privacy promise has a name behind it.

  • 🔒 Runs on your machine — words never leave
  • 🤖 On-device LLM inference
  • 🛠️ Free app + done-for-you setup
Academic foundation

The theoretical underpinnings.

Master of Science GPA 3.87

Customer Intelligence & Analytics

Pace University · 2018–2019

  • Microsoft Power BI
  • Data Analysis
  • Predictive Modeling
  • Strategic Intelligence
Bachelor of Science GPA 3.84

Finance & Computer Science

Pace University · Lubin School of Business · 2013–2017

  • Financial Analysis
  • Algorithms
  • System Design
  • Computational Finance
Credentials

Verified across the Microsoft stack and data science.

  • 🏆
    Power BI & Power PlatformMicrosoft
  • ⚙️
    Power Platform — Custom SolutionsMicrosoft
  • 🤖
    Copilot for Power BIMicrosoft
  • 💬
    Power Virtual AgentsMicrosoft
  • ☁️
    Azure SQLMicrosoft
  • 📈
    Power BI Data AnalystMicrosoft
  • 🛡️
    Trustworthy Generative AIVanderbilt
  • 🐍
    Python for Data AnalysisCoursera
  • 📊
    SQL for Data ScienceUC Davis

Want this on your computer?

Garnet is free to download and yours to keep. If you'd rather not fuss with setup, the engineer behind it will install it on your own computer for you — pick the right model for your machine, make sure it runs smoothly, and walk you through your first questions.

Connect on LinkedIn ↗ Book a setup →
Design signature · 2026

"Every part of Garnet is engineered the same way: small, sharp, owned. The same hands that built the private AI built this page — and the demo on it runs on the same kind of models."

— Jakub Rezayev, Founder & Principal Engineer