Eli Ferres

New York

Forward deployed systems engineer.
AI systems, operations and product.

Team player, builder and achiever by nature; excellence is my standard. Eight years at startups, teams of 50+, 2B+ organic views. I build systems, run teams and get 1 + 1 = 3 results across product, marketing, technology and operations, fully AI native.

Say helloSee the work

The work

Things I built, and how they work.

01

The personal or company OS

An AI assistant is only useful if it has the context. Scattered files and forgotten decisions leave it starting over.

Role and constraint
I built a shared brain and the workflows around it myself. Context stays backed up, and consequential actions need a person's approval.
The trade-off
Workflows move from shadow to draft to live. Code enforces the gates. An unanswered approval request expires.
Outcome
20,000+ logged checks; 30+ enforcement gates; 70+ tests, run nightly
Ferres Technologies, 2026See the repos on GitHub

The architecture

The personal or company OS

The personal or company OSEvery door loads one constitution and the gates in code, which write to one log; the gates read and write one vault, an always-on server runs the body, and any consequential action waits on the phone for a person’s yes.DOORSGATES IN CODE 30+BRAIN AND BODYCODE DOOREVERY ACCOUNTCODE DOORSAME GATESOTHER MODELSNOT AUTHORITYPHONE CARDVOICE NOTE,PHOTO, YESTHE CONSTITUTIONLOADED AT STARTHOOKSBEFORE EVERY CALL...AUDIT LOG 20,000+EVERY GATE WRITESPUSH WALLRECEIPT PER PUSHNIGHTLY TESTS 70+THE VAULTONE INDEX...SKILLS, ROUTINESOWN TRIGGERS...CLOUD SERVERALWAYS ONA PERSON’S YESUNANSWERED,IT EXPIRESWORLD STATEOPEN ITEMS, DOORS, ROUTINESREAD, WRITEREQUESTWRITESVOICE NOTES, PHOTOSWAITS FOR THE YES

One brain holds the context; every workflow reaches it through gates in code, and a person is the last step.

02

The market dashboard

The information is available. Reading the market still means working through dense dashboards and scattered feeds.

Role and constraint
I designed and built the dashboard myself, from data pipeline to screen, to help a person read the market.
The trade-off
When the paid feed died, I rebuilt around independent public sources. I test the engine behind each market read and put limits on how long a failing feed can hold it up.
Outcome
296 tests passing; 9 data sources; a market read every 60 seconds during market hours
Ferres Technologies, 2026

The architecture

The market dashboard

The market dashboardNine feeds pass a timeout and a breaker into one snapshot a minute, then a verdict engine of two lenses and factor slots; a board is built once a minute to the screen, a scorecard walks forward into the engine, and a person holds the kill switch.FEEDSPRICESFILINGSMACROSENTIMENTPREDICTION MARKETSTRIALSSHORT INTERESTOPTIONSNEWSTIMEOUT + BREAKERSNAPSHOTONE COPYPER MINUTEVERDICT ENGINEFACTOR SLOTS...ONE VERDICT PER TICKERFILINGSCATALYSTWATCHKILL SWITCHTHEBOARD60 SECONDSSCREENSCORECARDWALK-FORWARDTESTS 296

Nine feeds, one snapshot a minute, one verdict per ticker; the person holds the kill switch.

03

The operating platform for a multi-brand agency

Billing, cash, contracts and sales entry ran across spreadsheets and chat. The team needed a place to run the operation together.

Role and constraint
Head of product design and operations, and the platform's sole architect and builder. I ran every one of those processes by hand first, then built the internal platform myself: the creator-deal pipeline, a brand communication desk and the operations team's own board.
The trade-off
I moved the work in stages, kept the old records intact and made who sees the money a server decision. The team kept working through the change.
Outcome
18 modules, 4,700+ tests; every migration cut over in one day, nothing lost
Under NDA · 2024 to 2026

The architecture

Under NDA

The operating platform for a multi-brand agency

The operating platform for a multi-brand agencyRoles reach the app through one server that decides who sees the money; the modules and the scheduled jobs read and write one system of record beside a second store and feeds that only read in, and the move runs from frozen old records through waves to one day.THE APPROLESLAPTOP, PHONEDESKTOP FIRSTTESTS4,700+SIGN INSESSIONTHE SERVERWHO SEES THE MONEY,DECIDED ON THE SERVERMODULES...18 MODULESSCHEDULED JOBS...CASH, SALES, REPORTSSYSTEM OF RECORDBILLING, CASH,CONTRACTS, SALESSECOND STOREFEEDS, READ ONLYREAD / WRITEWRITEPULLTHE MOVEOLD RECORDSFROZEN, NEVER DELETEDWAVESONE DAY

The modules, at the level I can show: who sees the money is decided on the server, and the old records were never deleted.

04

The creator content engine

Eight YouTube channels and 50+ contractors ran without one production system.

Role and constraint
Director of Operations, Content, at ProGuides; Managing Producer at State Space after the acquisition. I built and ran the bench of 50+ contractors across all eight channels, and put one production loop under every channel.
The trade-off
I built the system and the team together, holding a 3:1 conversion-to-cost ratio as output grew.
Outcome
Views 500K to 2B+, followers 20K to 4M+
ProGuides / State Space, 2018 to 2023

The architecture

The creator content engine

The creator content engineA production wheel turning from brief through script, capture, edit, thumbnail and QA, publish and analytics; the team and the contractor pool sit inside it, the channels beside it, and the conversion-to-cost ratio on a gauge.BRIEFSCRIPTCAPTUREEDITTHUMBNAIL AND QAPUBLISHANALYTICSTHE TEAMWRITERS, EDITORS,ARTISTS, PRODUCERSCONTRACTOR POOL 50+CHANNELSVIEWS 500K > 2B+FOLLOWERS 20K > 4M+CONVERSION TO COST 3:1

A production loop the team turns together, holding the ratio as the channels grew.

What I bring

Five things I do well, with the receipts.

Product and design

I prototype first, against best-in-class references: the flow, the copy and the bar, before anything goes to production. Fully fluent in the product and design workflow.

Marketing and growth

I work across organic, paid and creator commerce. I took a gaming media company's short-form channel to 200M+ views, with views and conversions up 1,000 percent, and its training app from 5M to 20M+ monthly users.

Operations and finance

I have owned billing, cash reconciliation, contracts, capacity planning and a monthly profit-and-loss, and I fixed the process by hand before I automated any of it.

Systems and engineering

I build through AI systems and workflow orchestration. I direct the code, testing and delivery rather than write code by hand. This year I shipped two production apps and an agent layer myself, on top of the internal platform I built for an agency.

People and hiring

I made the first five hires at a startup and took it from idea to go-to-market with the CEO. Before that I built a fifty-person contractor bench across eight channels. I hire for hunger and judge by the portfolio.

Fluent in

The tools I use. The habits I keep.

Build

TypeScript, Next.js, React, Postgres, Prisma, Supabase, Redis, Stripe, Twilio, n8n, Vercel, GitHub, Claude Code

Run

Google Workspace, Slack, Zoom, Notion, Linear, Asana, Monday.com, DocuSign, Mercury, QuickBooks, Discord, Telegram, Obsidian

Reach

TikTok, YouTube, Instagram, Meta, Shopify and the creator-commerce tools

Think

Claude, GPT, Grok, Higgsfield, 21st.dev, Mobbin, Figma, Whisper, ElevenLabs

I ask for a named reference, never an adjective.

I bring a working prototype to the meeting.

I ship the system, then the doc.

I call it honestly. My own work too.

I measure it, or it did not happen.

Say hello

GitHubLinkedIn
Say hello

Say hello

Twenty minutes, at a time that suits.

A video call. Times are shown in your own time zone.

Prefer email? eliyahuferres@gmail.com