Skip to main content
Tailor Nate.
AI Engineer · Multi-Agent Systems

Nathan Hai
Vishnevski.

I deliver AI systems that ship.

I don't wait to be told where the value is.

IsraelCS @ HITMulti-agent systems
See the work
Open to work
02 / EVIDENCE

The Work

Used daily. Not demo-ware.

~/automates · main● live

AutoMates

My multi-agent AI development method.

Specialized AI agents in a structured pipeline (spec, plan, build, verify), with knowledge-first protocols, cross-session memory, and automated quality gates. Started as my own workflow problem and keeps evolving as the field moves: v1 went open source, v2 became my private working method. It's the factory that ships every other project on this page.

used daily · v1 open source · v2 is my private working method
stack
Claude CodeTypeScriptMCPMulti-AgentContext Engineering
automate-s.comgithub.com/Nate-Vish/Auto-Mates
~/polaris · ibm-israel● shipped

Polaris

Finance insight tool, demoed to IBM Israel leadership.

A dashboard that reads raw Excel financials and writes the story behind the numbers. One rule I refused to break: code computes every figure, AI writes text only. Nobody presents a hallucinated number to finance people.

MVP in one sprint · presented to IBM Israel leadership at demo day
stack
Gemini APINext.jsSheetJSVercelAutoMates
~/bet-hadar · production● shipped

Bet Hadar Portal

Procedures portal for a medical center.

Their procedures index lived in one shared Excel file that locked whenever two people opened it. I replaced it with an intranet portal, now live in production: instant Hebrew search, a read-tracking audit trail with Excel export, manager roles. I deployed it myself on their Windows Server (IIS, Active Directory, firewall) and trained the staff with an interactive Hebrew training module.

live in production · deployed on-prem by me · staff trained
stack
ASP.NET CoreC#Hebrew RTLIISAutoMates
full case study
~/los · first_step● in development

LOS

Agentic legal operating system for lawyers.

Building with a founding team: a Hebrew-first team of seven AI agents that does a small law firm's routine work end-to-end: drafting, research, review, office ops. AI drafts, the lawyer decides. Hard legal-safety rules mean the AI never pretends to be the lawyer.

in development with a founding team · 7 AI agents
stack
Next.jsTypeScriptMulti-AgentHebrew RTLAutoMates
~/madko · main● live

Madko

The task manager this site runs on.

A full-stack task-management product that runs my actual life, with an AI layer for task analysis, voice capture, a calendar feed and gamified XP. You're looking at it right now: tailornate.com is Madko's home. Next milestone: accounts, multi-device sync, web-push reminders.

live at tailornate.com · runs my day, every day
stack
Next.js 16TypeScriptVercel AI SDKPWAZustand
tailornate.com · you are here
03 / TOOLKIT

Stack

AI systems and data-driven decisions. Sharpest tools first.

Multi-Agent OrchestrationContext EngineeringLLMsRAGMCPPrompt EngineeringData-Driven DecisionsReal-Time IntelligenceClaude CodeTypeScript

AI & Agents

  • Multi-Agent Orchestration
  • Context Engineering
  • LLMs
  • RAG
  • MCP
  • Prompt Engineering
  • Agent Memory Systems
  • Automated Quality Gates

Data & Decisions

  • Data-Driven Decision Making
  • Real-Time Intelligence
  • Data Fusion & Verification
  • pandas
  • SheetJS
  • Zod

AI Tooling

  • Claude Code
  • Gemini CLI
  • Gemini API
  • Vercel AI SDK
  • n8n

Languages

  • TypeScript
  • Python
  • Java
  • C / C++
  • SQL

Web & Infrastructure

  • Next.js
  • React
  • Git
  • GitHub
  • Vercel
  • shadcn/ui
  • Linux
04 / OPERATING SYSTEM

How I work

For four years my job was walking into other people's worlds: eighteen civilian communities, each with its own people, its own fears, its own map. My job was to build the bridge between what they needed and what the system could deliver. I didn't know the title for it back then. Turns out it's the job description I've been living ever since.

Today I ship systems into places software rarely reaches gracefully: an IBM boardroom, a medical center's intranet, a Hebrew-speaking law office. AI is my specialty and my toolbox, but the point is always the delivery. I build the things I use: AutoMates started as my own workflow problem and is now the factory that ships everything else on this page.

Same year: full CS course load, a 200-day intelligence deployment, Polaris delivered to IBM Israel. That's not a hero story, it's how I'm wired. Depth when there's time, speed when there isn't. And any process that repeats twice becomes code the third time.

Hebrew · English · Russian. Off the clock: salsa, guitar, skiing, good coffee.

Operating principles
  1. Root, not symptom.Go to the root of the problem. Patches are interest on debt you'll pay later.
  2. Build a system.If a process repeats twice, the third time it's code.
  3. Plan before executing.For anything that matters. Not impatient on the things worth doing right.
  4. Obsessive when it clicks.Go all-in until it ships.
05 / SERVICE

Under pressure

Real-time, high-stakes. Where the delivery instinct was forged.

Nov 2024 - Present

Target Intelligence Officer, Reserves

Israel Defense Forces
  • Built automation workflows fusing multi-database operational data into decision-ready intelligence
  • Real-time act-or-wait calls in time-critical windows, under full operational responsibility
  • 200 days deployed while completing a full CS course load
Nov 2019 - Dec 2023

Territorial Defense Officer

Israel Defense Forces
  • Embedded with 18 civilian communities, learning each one's world and bridging it to army systems and resources
  • Trained and commanded 50 soldiers
  • Weekly briefings reaching 1,200+ soldiers
06 / CREDENTIALS

On record

B.Sc. Computer Science

HIT, Holon Institute of Technology
Currently 2nd year
2024 - 2027 (expected)

AI Driven Development Program

Future HIT
Accepted a year early. The program targets 3rd year+.
2026

Handzone Leaders Program

Future HIT · IBM Israel
Where Polaris was built.
2026

Target Intelligence Professional

IDF Intelligence Certification
2025

3rd Place, GenAI Hackathon

HIT
Nov 2025