Skip to content
Louay Elbiche
AI CONSULTANT, NEW YORK CITY

LOUAY
ELBICHE

Technical Product Manager. Full-stack AI Builder. AI Agent Orchestration. Forward Deployed Engineer.

I build and orchestrate AI systems that work autonomously.

I take embedded roles and engagements as a contract through Runwell Systems. I am based in New York City, available for remote work, and travel to client sites on demand.

See the workGet in touch
runwell
$ runwell status
ROLESexplicit ownership, delegation rules
LEARNINGcorrections become standing rules
EVALSagents evaluated before they ship
GATESfunction, performance, security
$ runwell sessions
command center: handoff proven
validated until clean
$
FOUR ROLES, ONE PRACTICE

WHERE I FIT

I design, build, ship and operate AI systems for my clients' goals: one senior operator from architecture to production.

TECHNICAL PRODUCT MANAGER

Deep technical work with engineers, before anything is built.

Seven years in product management, 2018 to 2025, across music streaming, mobile, fintech and blockchain infrastructure. At Blockdaemon, Influence Mobile and Spryte Labs I worked with engineers on a deep technical level before anything was built, and owned the roadmap.

The results

  • 70% shorter enterprise sales cycle
  • 20% more sales
  • 90% less demo complexity
  • 30+ blockchain protocols onboarded
  • 50% lower infrastructure cost
Read more: Technical Product Manager 

FULL-STACK AI BUILDER

Everything we build at Runwell Systems, I build end to end.

A multi-tenant business platform with an AI concierge, a bilingual training and learning platform with an AI tutor, our own sales and CRM platform, a universal web scraper, a Shopify AI shopping assistant, a job market intelligence platform, a fully end-to-end automated tax tool for CPAs, desktop and iOS apps, an inventory ERP with a scan gate, and brand identities from scratch.

16 client projects built for businesses, with 9 client websites and platforms delivered. AI we run: our own agent orchestration system, and AI products: a multi-tenant chatbot platform, an AI tutor inside a learning platform, and Creative Studio.

See the work 

AI AGENT ORCHESTRATION

The system that runs Runwell Systems, ready to apply to your operations.

I designed and run a multi-agent orchestration system that operates my own company: specialised agents with explicit ownership boundaries, delegation rules and a written decision record for every structural choice. I can apply the same design at your company, at any scale, to any need, with every agent evaluated before it ships. Automation of the hand-offs between its parallel work sessions is in development.

Security is built into how it runs. Every change passes a chain of blocking gates (a secret scan, static analysis, a container scan and a dependency audit), each agent works inside its own tool permissions, and AI code stays in private repositories.

See what I would build 

FORWARD DEPLOYED ENGINEER

Embedded in your business: diagnose first, then build, ship and run.

This is the role and the term for how I work inside a client's business. I enter unfamiliar businesses and diagnose before building: discovery and a written plan came first for a cross-border CPA advisory, a boutique hotel and a medical supplies company. Then I build, ship and run the system, with staged releases, per-tenant health monitoring with alerting and governance built in. Client systems I delivered are in production for businesses in tax advisory, hospitality, medical supplies and e-commerce.

Client systems get the same security discipline: the gate chain before every release, tenant isolation on multi-tenant platforms, and AI code kept in private repositories.

See the client work 
OUTCOMES

Client work at Runwell Systems, and the product record behind it.

RUNWELL SYSTEMS, CLIENT WORK

  • 16 client projects built for businesses.

  • AI we run

    Our own agent orchestration system, and AI products: a multi-tenant chatbot platform, an AI tutor inside a learning platform, and Creative Studio.

PRODUCT CAREER, CORPORATE

  • Seven years in product management, 2018 to 2025, across music streaming, mobile, fintech and blockchain infrastructure.

  • 70% shorter enterprise sales cycle for an institutional cryptographic key-management product at a blockchain infrastructure company, through an AWS Marketplace deployment I architected.

  • Sales up 20% in six months after I launched self-checkout, pay-as-you-go purchasing for the same key-management product.

  • A sandbox environment I designed so enterprise prospects could test the key-management product's key features, cutting demo complexity by 90% and freeing engineering time for implementation.

  • Onboarded across 30+ blockchain protocols to automated validator node deployment and staking workflows under one API, with every protocol in a node management dashboard that unlocked self-service onboarding.

  • Blockchain nodes migrated from cloud to bare metal at a blockchain infrastructure company, cutting infrastructure cost by 50%, with server provisioning and lifecycle management automated.

  • A no-code website builder I launched for an ETF firm in asset management: 10x faster time to market and 5x client capacity.

  • A refactor of a main client-facing mobile app for cross-platform deployment, with a 50% faster release cycle.

ACROSS MY WHOLE RECORD

  • 12 industries served, across my product career and my client work: VR and AR; blockchain infrastructure; fintech and financial services; healthcare and medical tourism; hospitality and wellness; legal and education; medical supplies; mobile apps; music streaming; professional services; retail and commerce; scientific and industrial b2b.

SELECTED WORK

The systems I built, and what each one does.

01

A MULTI-TENANT BUSINESS PLATFORM WITH AN AI CONCIERGE

A platform that gives each business its own website, ordering, booking and store, with an AI concierge on its site and messaging channels that answers customers, captures leads and takes bookings.

  • A command line scaffolds a new business on the platform, and the messaging gateway routes every conversation to the right tenant.
  • Owners run leads, bookings, staff and inventory from one back office, with an AI advisor beside them.
  • Website chatbots grounded in each business's own content are live on client sites, served from this platform.
02

A BILINGUAL TRAINING AND LEARNING PLATFORM WITH AN AI TUTOR

A bilingual training marketplace and learning platform, built end to end, with an AI tutor on every course.

  • The tutor answers only from the course's own content, with citations, declines out-of-scope questions and hands the learner to a human.
  • Teacher tools generate study guides, flashcards, mind maps and multiple-choice quizzes; the teacher validates them, and validated quizzes go into the course's quiz engine.
  • Access is granted per course, the tutor runs under hard cost and token caps, and a public rate-limited demo is open to visitors.
03

A SALES AND CRM PLATFORM

Our own sales platform: lead discovery, a CRM pipeline, go-to-market planning, market intelligence, prospect documents, communications, demos and tasks in one place.

  • AI qualifies leads, assesses a business and drafts the outreach message.
  • Contacts are grouped into segments and partner lists, and a cost ledger tracks what outreach spends.
  • It runs our own sales operation every day.
04

A UNIVERSAL WEB SCRAPER

A scraper that reads any site, API or browser-rendered page, runs full crawls, and uses AI to extract the data into a schema you define.

  • It respects robots.txt and rate limits, and a stealth browser retries through bot walls.
  • It is served as an API for scraping, streaming, preview, lead discovery, contacts, reviews and ad libraries, and as an MCP server so AI agents can call it.
  • A stress test of about 2,500 requests returned no errors, at 330 to 380 requests per second on the fast paths.
05

A SHOPIFY AI SHOPPING ASSISTANT

An AI shopping assistant that understands each store's catalog through vector search and helps its customers, on merchant stores.

  • It can search products, show details, check stock, add to cart and report order status.
  • It detects what the shopper wants, follows the session, suggests upsells and personalises results.
  • Merchants get a dashboard.
06

A JOB MARKET INTELLIGENCE PLATFORM

A platform that ingests job postings, with each source's terms respected, and turns them into labour-market insight.

  • Postings are mapped to the ESCO and ROME occupation frameworks.
  • It adds analytics and AI search over the postings.
07

A FULLY END-TO-END AUTOMATED TAX TOOL FOR CPAs

A fully end-to-end automated tax tool for CPAs that prepares federal and state returns.

  • It covers Schedule C, Form 1040, state returns, amendments and education credits.
  • It tracks documents, traces income and fills the PDF forms, with its rules checked against IRS sources.
  • It flags items for CPA review and handles many clients in one place.
08

DESKTOP, IOS AND WEB APPS

A desktop app, an iOS app and web apps, built for our own products.

  • The desktop and iOS apps share one TypeScript core, with local SQLite storage.
  • A private video-feed web app runs behind a login.
09

AN INVENTORY ERP WITH A SCAN GATE

A website and an ERP inventory management system for a medical supplies distributor, built after discovery and a written plan, then handed over for deployment and maintenance.

  • Its intelligence layer runs analysts for sales, margins, trends and reorders, and an inventory advisor answers business questions in chat; every change lands in a tamper-evident audit trail.
  • At the gate, two OCR engines vote on each reading, GS1 barcodes are scanned, and every delivery is checked against its delivery slip.
10

MORE SYSTEMS

  • A tax advisory chatbot for a CPA firm, with answers grounded in IRS tax law through retrieval (RAG): hybrid keyword and semantic search, re-ranking and a source citation in every answer.
  • A full storefront for a retail business, with a lead capture form and an AI chatbot grounded in the business's own content.
  • A company website for a professional services firm, built mainly for discoverability, with an SEO audit completed and blog posts published.
  • A full medical tourism platform, with its front end built.
  • A knowledge extractor that turns video into transcripts, on-screen text and a searchable knowledge base.
  • A discoverability tool that audits a site for search engines and for citation by AI assistants.
  • AI creative production for marketing and advertising: image, video and brand-asset pipelines that give campaigns more creatives to test, and only the promoted winners run.
  • An organic marketing program for a US research supplier that cannot buy paid ads: search and AI-assistant discoverability, research-backed social content and a newsletter, with every study claim cited to its source.
11

THE ORCHESTRATION SYSTEM BEHIND RUNWELL SYSTEMS

A multi-agent orchestration system that operates my own company: specialised agents with explicit ownership boundaries, delegation rules and a written decision record for every structural choice, coordinating with each other and with me.

  • It learns from its own work: failures and corrections are captured and turned into standing rules the agents follow.
  • Agents are evaluated before they ship: end-to-end QA suites run against production, and a grounding gate blocks any unsourced claim from shipping.
  • Multi-session orchestration: parallel AI work sessions under a command-center session, which messages each of them, hands off work only after proving it was picked up, and runs fresh validation rounds until clean.
  • Separate release gates for function, performance and security, including LLM-safety probing, run by QA agents that did not build the work.
  • Every change passes a chain of blocking gates: a secret scan, static analysis, a container scan and a dependency audit.
  • Claims computed from records: which clients and products may be named is derived from the registries by validators, not typed by hand.
  • Knowledge graphs for agents: code and schema graphs agents cite instead of searching blind, and a relationship-aware marketing knowledge graph.
  • A generated map of the whole platform (products, registries, lifecycle states and gates), rebuilt from the source of truth and drift-checked so it never goes stale.
The control loop: the same seven steps on every request, from a plain-English ask to shipped, remembered software.

  1. A request, in plain English. Typed in any project directory, or fired on a schedule. No ticket, no spec, no syntax to learn.

  2. A router classifies it. It names the discipline that owns the work before any work begins, so every task has one owner and a traceable path back to the request that caused it.

  3. The orchestrator assigns it. The single brain. It does no specialist work itself and it is the only thing permitted to start an agent, so control stays hierarchical. Heavy or parallel work gets its own agent with its own context; focused judgment is adopted in place.

  4. A specialist takes it. A rulebook rather than a prompt: a defined scope, hard gates it cannot exit without passing, and its own tool permissions. It loads shared skills on the way, so a deploy or a security review runs the same way whoever invokes it.

  5. It is implemented. Written to the conventions already in the codebase: same languages, same architecture, same style, so what comes out is code the team would have written itself. Values come from registries that hold the facts, derived rather than typed.

    • TypeScript
    • Node
    • Python
    • PostgreSQL
    • React
    • Next.js

  6. Every claim is checked. Secrets, static analysis, dependencies, tests, review. Each check is bound to the thing that generates the claim, and blocks rather than warns. A gate that can be walked around is decoration.

  7. It ships, and is remembered. Feature branch, atomic commit, pull request, squash merge, then verification against the running surface. What was learned is written to persistent memory, so the next request starts further along.

WHAT RUNWELL SYSTEMS DOES

We build and run websites for client businesses, and the AI platform that runs alongside them.

Our chatbot platform gives a business an AI chatbot that answers customers, captures leads and takes bookings, with a back office where owners manage leads, bookings and staff.

We also build AI creative production pipelines that make images, video and brand assets for marketing and ad campaigns, so campaigns get more creatives to test.

Before we build, we start with discovery and a written plan, as we did for a cross-border CPA advisory, a boutique hotel and a medical supplies company.

Our client systems are in production for businesses in tax advisory, hospitality, medical supplies and e-commerce.

HOW I WORK

DIAGNOSE FIRST, EVALUATE BEFORE SHIPPING

  1. 01

    I enter unfamiliar businesses and diagnose before building: discovery and a written plan came first for a cross-border CPA advisory, a boutique hotel and a medical supplies company.

  2. 02

    Agents are evaluated before they ship: end-to-end QA suites run against production, and a grounding gate blocks any unsourced claim from shipping.

  3. 03

    Governance is built in: encrypted secrets with staged rotation, commit-time leak prevention, staging gates and audit trails.

  4. 04

    Production discipline: staged releases, per-tenant health monitoring with alerting, migration validation and registries that record failures honestly.

  5. 05

    I say no to the wrong AI work: I retired an offering with no market and qualify out engagements that do not fit, rather than overpromising.

ABOUT
Louay Elbiche

I spent seven years in product management, across music streaming, mobile, fintech and blockchain infrastructure. At a blockchain infrastructure company I owned an institutional cryptographic key-management product, and the institutional wallet and staking products. In November 2025 I founded Runwell Systems, and I run it on a multi-agent system I designed. I work across product and engineering in one seat. I am based in New York City, available for remote work, and I travel to client sites on demand.

I work inside a client's business to find where AI and automation can pay off, then build, ship and run those systems toward the goals they set.

EXPERIENCE

  1. Founder, Runwell Systems, from November 2025.

    • 16 client projects built for businesses.
    • I designed and run the multi-agent orchestration system that operates the company.
    runwellsystems.com 

  2. Technical Product Manager, Blockdaemon, November 2021 to March 2025.

    • Owned the node lifecycle and deployment product and its roadmap, an institutional cryptographic key-management product, and the institutional wallet and staking products.
    • On the node product, worked daily with blockchain engineers and interface developers on integration points and requirements, ran sprint planning and backlog refinement, and aligned product vision, revenue and compliance objectives with the engineering plan.
    • Wrote requirements and technical solutions with engineering against the real stack (Ansible, Kubernetes, DNS services and cloud provider APIs), and designed the workflows that handle private data such as wallet secrets and sensitive node configuration.
    • Roadmap items included node migration and protocol onboarding, all nodes and all customers in the app with invite-link onboarding, self-deployed full nodes, high-availability RPC as a service and clusters, customer node configuration in the web app, and monitoring stack requirements.
    • Architected the AWS Marketplace deployment of the key-management product, cutting the enterprise sales cycle by 70%.
    • Launched self-checkout, pay-as-you-go purchasing for the key-management product; sales up 20% in six months.
    • Designed a sandbox environment for prospects to test the key-management product's key features, cutting demo complexity by 90% and reallocating engineering time to implementation.
    • On the institutional wallet and staking products, owned the journey from requirements through testing to unify API and wallet onboarding into one product journey.
    • Onboarded 30+ blockchain protocols to automated validator node deployment and staking workflows under one API, and integrated them all into the node management dashboard, unlocking self-service onboarding.
    • Migrated blockchain nodes from cloud to bare metal, cutting infrastructure cost by 50%; automated server provisioning and lifecycle management.

  3. Technical Product Manager, Influence Mobile, March to November 2021.

    • Led the redesign of the main client-facing app in React Native for cross-platform release, managing the developers, roadmap, sprints and technical requirements, for a 50% faster release cycle.
    • Built a data strategy on Firebase: chose the events to track, analysed audiences to find churn and weak points in the experience, and ran A/B experiments to find the behaviour worth encouraging.
    • Built a crypto rewards adoption plan with third-party APIs, opening a new user acquisition channel.

  4. Product Manager, Spryte Labs, June 2019 to March 2021.

    • Led the design, development and launch of Webmaker, a no-code website builder for an ETF firm in asset management, with 20+ ETF websites launched on it.
    • Ran a team of five developers in Scrum; architected the modules, data pipelines and APIs on a headless CMS over PostgreSQL; set up automated builds and deployments (Jenkins, Docker); designed the interface and dashboards in Figma; and was hands-on in Vue.js and Python.
    • 10x faster time to market and 5x client capacity.

  5. Associate Product Manager, SiriusXM, June 2018 to May 2019.

    • Drove the integration of a music streaming service's backend into the company's app to deliver Artist Radio Stations; user metrics analysis, UX optimisation and bug triage.
SKILLS
EDUCATION
FAQ
What does Louay Elbiche do?

He is an AI consultant: he designs, builds, ships and operates AI systems for his clients' goals, as one senior operator from architecture to production.

Where is Louay based?

New York City. He is available for remote work and travels to client sites on demand.

What is his experience with AI agents?

He designed and runs a multi-agent orchestration system that operates his own company, with explicit ownership boundaries, delegation rules and a written decision record for every structural choice. The system learns from its own work, and agents are evaluated before they ship.

What has he built for clients?

A multi-tenant business platform with an AI concierge on each tenant's site, a bilingual training and learning platform with an AI tutor, a sales and CRM platform, a universal web scraper, a Shopify AI shopping assistant, a fully end-to-end automated tax tool for CPAs, desktop and iOS apps, an inventory ERP with a scan gate, and a tax advisory chatbot grounded in IRS tax law for a CPA firm. Client systems he delivered are in production for businesses in tax advisory, hospitality, medical supplies and e-commerce.

How many client projects does Runwell Systems have?

Runwell Systems has 16 client projects built for businesses. AI we run: our own agent orchestration system, and AI products: a multi-tenant chatbot platform, an AI tutor inside a learning platform, and Creative Studio.

What is his product management background?

Seven years in product management, 2018 to 2025, across music streaming, mobile, fintech and blockchain infrastructure. At a blockchain infrastructure company he owned the node lifecycle product and an institutional key-management product, and architected the AWS Marketplace deployment that cut its enterprise sales cycle by 70%.

What is his software engineering background?

He holds a B.S. in Computer Science from Florida Institute of Technology and an M.S. in Information Systems from Stevens Institute of Technology. He is hands-on in engineering: he architected the backend and CI/CD (Jenkins, Docker) of a no-code website builder for an ETF firm in asset management and was hands-on in Vue.js and Python, and he now designs, builds and operates production AI systems in TypeScript, Python, Next.js, React, FastAPI and PostgreSQL.

What is Runwell Systems?

Balencia LLC d/b/a Runwell Systems, founded by Louay Elbiche in November 2025.

Which technologies does he work in?

TypeScript, Python, JavaScript; Next.js, React, FastAPI; PostgreSQL, pgvector, SQLite, Redis, Prisma; Docker, nginx, PM2, AWS, CI/CD; Claude API, Gemini, OpenAI embeddings, RAG pipelines, vector search.

How does he decide what AI work to take?

He diagnoses before building, and he says no to the wrong AI work rather than overpromising.

Which languages does he work in?

English, French and Spanish.

How does he handle security?

Every change passes a chain of blocking gates: a secret scan, static analysis, a container scan and a dependency audit. AI code stays in private repositories, multi-tenant platforms isolate each tenant, and each agent works inside its own tool permissions.

GET IN TOUCH

LET'S TALK.

Tell me about the role, the engagement or the problem.

or email me directly