Eddy R. Lora

Eddy reading the back of a book between the shelves of a bookstore, in black and white.

Eddy at night beside a lamp post in a park, in black and white.

Eddy lining up a shot on a pool table in a bar, in black and white, a friend waiting with a cue.

About.

Before Rutgers I went to Woodberry Forest in Virginia, where I played football and basketball, ran track and read in chapel. At Rutgers I studied English with a minor in history and made the Dean's List.

At Woodberry is where I started to pick up on desires of mine, like debate and literature. I didn't know it until I graduated, but being part of a tight-knit community, you're exposed to so much conversation, conversation that doesn't seem like much at first but is constantly challenging your beliefs. At Rutgers, in the English program, I met many people and professors who thought this way. The classes were small and everyone took part in the discussion. Constantly analyzing and brewing creativity is what I enjoy, and what I try to carry into my work.

A note is an argument, the model is the evidence, and the writing is where a reader decides whether to trust either.

I interned on two research desks at the same time while I was a full-time student, finished first in my region at TD Bank on twenty to thirty calls a day, co-founded a marketing software company and turned it into my own agency, and came back to research with the systems I'd built along the way. I passed the Securities Industry Essentials in September 2026.

Away from work, chess most days, basketball since AAU, weightlifting, golf when someone has a tee time, pool when there's a table, and building my own PCs. The Catcher in the Rye is my favorite book of all time, then This Side of Paradise. The ones I keep going back to for work are The Disciplined Trader, Man's Search for Meaning, Market Wizards and Meditations.

Rutgers, The State University of New JerseyBachelor of Arts in English, minor in History. Dean's List. May 2025.

Woodberry Forest School, VirginiaFootball, basketball, track. St. Andrews Chapel head reader. May 2021.

One note published, on Modine, and two more built, on TE Connectivity and Eaton. The other names in the frame are ones my trading engine scans.

Modine Manufacturing, an initiation.

July 27, 2026, 32 pages

A 32-page initiation on Modine, a company changing what it is, published two days before the company reported. The bear case mixed two sets of numbers. The $88.4 million of corporate cost it quotes is the GAAP figure, and on the adjusted basis the note uses it is $62.6 million.

Read the note

Page three of the Modine note, the key takeaways.

03

The cover of the Modine note, July 27, 2026.

01

What	Old	New	Why
Corporate cost in the high and bear cases	The GAAP $88.4 million set against adjusted numbers	$62.6 million, from the figures Modine re-issued on July 24	The same mix I argued the bear case made.
The margin the rule watches	Worked out by hand every time	A column in the model that computes 15.29 for the quarter	So the hand figure and the model always agree.
Data Centers growth for the quarter	A growth rate built from two different sets of segment figures	Removed, with the reason written in the cell	Better blank than wrong.
The full-year check for fiscal 2027	Compared one quarter to nothing	Switched off, with its reason written in the cell	So a skipped check can't read as a pass.

Four of the changes to the model after the first quarter.

My model.

I built the operating model in Excel, by hand and with Python. Python brings in the filings, so every historical cell is a filed number or a dated assumption. The forecast and the valuation are mine. Before the note goes out, every number it states is checked against a live cell.

After the first quarter I rolled the model forward and logged every change. Nine of nine reported lines matched within $0.05 million, the limit I set.

The first quarter after the note came in under 16.0 percent.Source: Modine's first-quarter fiscal 2027 release, July 29, 2026, and the segment history Modine re-issued on July 24, 2026.

The rule I set, and the first quarter.

I wrote the rule into the note before it went out. Segment margin under 16.0 percent for two quarters in a row changes the call.

The first quarter came in at 15.29 percent, the first quarter under 16.0 in the history Modine re-issued on July 24. The rule says two quarters, and it says two on purpose, so the call held.

The second quarter decides it.

Python and Claude.

I model the way a desk does, in Excel and by hand. Python and Claude Code are how I move faster as I learn. I use Claude to read the filings I pull, check the first cut of my model and read the draft back cold. I make the call.

I pull ten years of filings.
I build the model from them in Excel.
Every figure is checked back to its filing. One mismatch stops the build.
A second read, cold, tells me where an analyst would push back.
I read all of it, make the call and write my own sections.

The ask as I type it, the run, and the three notes so far.

My research experience.

Two desks at once in 2023, both while I was a full-time student at Rutgers.

C.L. King & Associates Hoboken, New Jersey. May to September 2023.

This is where I decided on equity research. I worked under a senior analyst with thirty years in the business. He was short-handed, and for a stretch I was the only person under him, so I learned fast.

I started by keeping up his models on his specialty chemicals and materials names. Then he gave me Rogers Corporation, a company he had never covered, and had me build the initiation on his template. The industry and the peers, the model's history and a forecast off it, calls with management and investor relations, the morning call. Much of it I worked out myself. It's the same process I ran on Modine, and now on Eaton and TE Connectivity.

Auerbach Grayson New York. April to October 2023.

Auerbach was the early shift. Research from partner brokers in Asia and Europe landed overnight, and before the US open I read every report, pulled the facts for each ticker together, checked them with the analyst and sent them out. It's where I learned to be on time every time, and to check before I send.

My trading experience.

I started in August 2022, when the Fed's tone shook the market and growth names sold off hard. I traded the semiconductor names down and back up through the whole turn, and the S&P beside them. Short trades at the open, off a fifteen-minute chart with the one-minute beside it. 128 closing trades in 2024, Nvidia and SPY the two heaviest.

My chart was plain. Trend lines, supply and demand, volume, a 50-day average, or a 9 on the shorter frames. The entry was always the close, never the touch.

Write it down, or it didn't happen. Every fill went into a journal of nineteen fields, entry and exit to the second, the exit tagged by its reason. A few rules never moved. Three to one. The exit set before the entry. Take every trade that meets the definition and none that doesn't. In October 2024 I wrote down a grade for each setup, from a plain setup at a line up to one with volume and a dated catalyst behind it. I sized against the grade, and I never checked whether the grades were real.

Source: daily bars read from my trading engine, November 2023 to July 2024. The bars are real daily prices and the ticks inside them are drawn.

Trade #
Date
Ticker
Entry Price
Exit Price
Strategy
Exit Signal
Strikes
Type
Time of Day
Entry Time
Exit Time
Duration (minutes)
Entry Mark
Exit Mark
Difference
three columns withheld

Source: the header of my own journal, September 2024, three columns withheld. The rows stay on my machine.

Testing my own rules.

In 2026 I rebuilt those rules in Python. One set of rules grades every setup the same way, in the history and in the live scanner, so the two can't drift apart.

Prices for every name, pulled and cleaned.
One set of rules grades every setup, in the history and live.
1,771 trades replayed across 156 names, 2010 to 2026.
The grades tested against 20,000 permutation and 20,000 block bootstrap draws.
The separation in the history doesn't hold in the years it never saw, so nothing is sized on it.

364tests the code runs, every one passing
156names the scanner reads
1,771trades replayed, 2010 to 2026
20,000permutation draws
20,000block bootstrap draws

Read from the engine when this page was built. If one test fails, the page does not build.

pytest

$ python -m pytest........................ [ 39%]........................ [ 59%]........................ [ 79%]........................ [ 98%].... [100%]364 passed in 114.21s (0:01:54)

The engine's last test run, read when this page was built.

Ten thousand shuffles of the grades in each window, and the gap between the top grade and the bottom that I actually measured, as a line. In the years I built on, it sits outside almost every shuffle. In the years I held out, it sits inside them.Source: the engine's own null run, 20,000 permutation and 20,000 block bootstrap draws, re-run read-only when this page was built and matched to its saved summary.

Reminous.

It started as software. In October 2024 I co-founded Magnifye, marketing software for restaurants, and when owners wanted the work done we started running the ads ourselves. My co-founder left, we dissolved it, and I kept going under my own company, Reminous.

Reminous runs ads for owners and builds what the ads need after the click, the page they land on, the lead system behind the form, the follow-up and the weekly report.

It's also where I learned to build systems that run without me, and that's the same skill I brought back to research.

reminous.com, built, not published yet

The whole Reminous site, top to bottom, scrolling by itself in a browser window.

How it runs, on the Reminous site: the calls the team makes by hand, and the reading AI does for them, each step marked done.

The weekly report on the Reminous site, a dashboard for a made-up week of ads for a made-up business.

The Reminous site as it stands. The client work on it is real. The example business is a sample and is labeled as one.

The systems behind it.

Each one below is a live workflow, read straight from n8n, the tool that runs them, and drawn from its own steps. They run on their own, every time.

n8n Lead Bus

Drag the drawing sideways to follow it.

Lead Bus. The site form's leads. Put in one format and checked for spam, then logged, and the office hears at once in business hours or at nine on the next business morning, and the person gets a text only past a consent check.Meta Lead Watcher. The ad form's leads, read off the sheet as they land, put in one format and sent to the same hours and the same consent gate.Boost Form Lead Bridge. A boosted post made its own form with nothing behind it, so this reads that form straight off the ad platform and writes each new lead where the watcher looks.Inbound SMS + Voice. Calls to the business number ring straight through. Text replies are checked as genuine, emailed to the office at any hour and texted on in its hours, and genuine calls are logged.Lead Heartbeat. Every three hours, the latest ad-form leads are matched against the alerts that went out, and anything missed pages me.Error Handler. When any workflow fails, it decides whether the failure is real, then tells me by email and by text.Spend Watchdog. The day's and the month's ad spend against the limit, every hour, and an alert when either crosses it.Weekly Report. Monday mornings, the week's ads against the week before and the lead ledger, written up and sent.Uptime Watch. The site's home, quote and contact pages and n8n itself, checked on a clock, and anything wrong tells me.SMS Delivery Watch. Every text that fails, caught from the carrier's own callback and from a sweep of recent messages, and reported.Twilio Balance Watch. Each morning the texting balance and last month's use become a runway, and a low balance texts and emails me.Twilio Usage Alert. The carrier's own usage alert, shaped into something readable and emailed.Consent Trail Export. On the first of the month, the consent trail is exported to a file and mailed, so the record never lives in one place.n8n Self-Backup. Every night every workflow is exported to a file and mailed, so none of this lives in one place either.

My App.

Coming soon

An arena and learning platform for Excel, Python and working with AI, the chess.com shape applied to finance skills. Play inside a real spreadsheet window or a terminal and get graded on your keystrokes, race a ghost of your best run or the day's shared board, then review the game key by key. One spreadsheet engine with 103 checks behind it, and nothing gets graded that the engine hasn't checked.

This is the fun one, and the thing I'd show first if you asked what I make for myself. Free to train and free to play, no account needed. It comes from the same place as the rest of this page, the desk skills drilled until they're automatic. Unnamed here until it's out.

The app's home screen, the three games, the course, the sprint, the daily puzzle and the model.

A model level, a spreadsheet window on Modine's filed figures with the move taught over it and the task beside it.

A year on GitHub.

github.com/EddyRLora

2,554

contributions in the last year, read from my GitHub profile when this page was built. Everything on this page is in there, in private repos that sync from my machine through the day.

Building with Claude.

7,400

prompts to Claude. Claude only keeps its own counts from March 28 for prompts and July 25 for tokens, so this starts there. Since July 25: 28 billion tokens.

EDDYR.LORA