Edgar Lopes Salvador
Download CV

Market analyst & futures trader · Tilburg, NL

I build AI systems that run marketing, research and coaching work.

I have traded FX and futures since 2020. In 2026 I started to build software with Claude Code. Eight of those systems are on this page. Each one has a demo you can run in your browser.

Portrait of Edgar Lopes Salvador
Edgar Lopes SalvadorMarketing operations & AI systems at Toodegrees
Try the systems Contact me
Activity across my systemssample data
    Each line is the kind of event a system logs while it works. Hover to pause.

    Eight systems

    Built between February and October 2026. Click a row to jump to its demo.

    Toodegrees OSAI workspace for a trading-education teamlive
    Social OS13 AI agents that plan, write and publish social postsin production
    SEO ConsoleWeekly SEO and AI-search report for four clientsruns weekly
    Check-in OSCoaching platform that turns check-ins into coach actionslive
    FX Strength TerminalRanks 8 currencies and suggests trade pairstrading tool
    Macro TerminalReads the US economic regime from 50 data seriestrading tool
    MTA MasteryTrading course with interactive chart lessons and certificateslive
    Psychology AssessmentMaps trader psychology and tracks it every monthlive

    Live · os.toodegrees.trade · internal tool

    Toodegrees OS

    The team picks a task. The app adds the brand voice, the SOP and good examples from a 756-file knowledge vault. Then Claude writes the draft and a person approves it.

    • 70pages and 63 API routes
    • 21agent specs, run every 5 minutes on a schedule
    • 115+automated test suites, 290+ commits since July 2026

    Next.js · TypeScript · SQLite · Claude · OpenAI · Ollama · Docker

    1. Pick an agent or template
    2. Librarian finds context in the vault
    3. Build the prompt: voice, SOP, format
    4. Claude writes the draft
    5. A person reviews it
    6. Publish or schedule
    toodegrees-os
    Open full screen ↗

    In production · Aligned Performance

    Social OS

    A social media team made of 13 AI agents. It decides what to post and when, writes the copy, publishes it, and learns from the results of every post.

    • 13specialist agents with one shared memory
    • 80+posts published without manual work
    • 12screens, about 16,500 lines of TypeScript

    Next.js · React · SQLite · Claude · Remotion · YouTube, Instagram and TikTok APIs

    1. Sync metrics from each platform
    2. Update what it has learned
    3. Run A/B experiments
    4. 13 agents plan the next post
    5. Publish, or ask for approval
    6. Save the lessons to the vault
    social-os
    Open full screen ↗

    Runs every Saturday · 4 clients

    SEO Console

    It checks whether ChatGPT, Perplexity, Gemini and Claude recommend each client. It logs every gap. Then Claude writes the weekly brief and the plan.

    • 11 minfor a full unattended weekly run
    • 1 linein a registry file adds a new client
    • 2languages: English and Dutch

    Node.js · PowerShell · Claude Code (headless) · DataForSEO · Rybbit · Search Console

    See the results in the SEO & GEO case study ↓

    1. Fix the buyer questions
    2. Ask 4 AI engines
    3. Log the citation gaps
    4. Collect the weekly metrics
    5. Claude writes the brief and plan
    6. Publish the report, scan again
    seo-console
    Open full screen ↗

    Live · os.alignedperformance.net

    Check-in OS

    Coaching clients log how they slept, how stressed they are and which habits they did. Rules turn each check-in into a protocol for the client or an alert for the coach.

    • 96check-ins created per program: 84 daily, 12 weekly
    • 5logic engines, 19 pages, about 7,600 lines
    • RLSeach client sees only their own data, enforced in the database

    React · TypeScript · Vite · Supabase · Cloudflare Pages · Resend

    1. Coach starts a 12-week program
    2. Client submits a check-in
    3. Trigger rules run
    4. Trends and risk flags update
    5. Matching resources appear
    6. Coach steps in
    checkin-os
    Open full screen ↗

    Trading tool · updates every 60 s

    FX Strength Terminal

    It scores the 8 major currencies on macro data, futures positioning and trend. It then pairs the strongest against the weakest and flags crowded positions.

    • 8currencies, 3 factor groups each
    • 10steps in the scoring engine
    • 9tabs, including oil, gold and indices

    Node.js · Express · server-sent events · FRED · CFTC · Alpha Vantage · TradingView

    1. Fetch data from 5+ sources
    2. Score each currency
    3. Quant engine: positioning, momentum
    4. Suggest pairs, raise alerts
    5. Push live to the dashboard
    fx-terminal
    Open full screen ↗

    Trading tool · Federal Reserve data

    Macro Terminal

    It reads 50 US data series and gives each indicator a green, amber or red signal. The signals combine into one call on the economic regime.

    • 50FRED series, each with commentary
    • 15futures contracts for positioning
    • 10signals combined into one regime call

    JavaScript · Node.js · Express · Chart.js · FRED API · CFTC API

    1. Fetch FRED and CFTC data
    2. Compute KPIs and the COT index
    3. Give each indicator a signal
    4. Combine into a regime
    5. Show it on an 11-tab dashboard
    macro-terminal
    Open full screen ↗

    Live · Mindful Trading Academy · course in Dutch

    MTA Mastery

    An online trading course. Each lesson explains a concept, shows it in an interactive chart and ends with a quiz. Students earn a certificate per course track.

    • 64Dutch lessons across 4 course tracks
    • 10interactive chart scenes, built with Three.js and canvas
    • Supabasemoved from Base44 as the platform grew

    React · Vite · Tailwind · Three.js · Supabase · jsPDF

    1. Student logs in
    2. Opens the learning map
    3. Lesson: concept, chart, scenario
    4. Quiz and reflection
    5. Progress is saved
    6. Certificate is issued
    mta-mastery
    Open full screen ↗

    Live · coach tool inside MTA

    Psychology Assessment

    Students answer an intake about how they behave as traders. A scoring engine maps them to 7 trader archetypes. Coaches then track the change with a monthly check-in.

    • 100intake questions, 20 per monthly check-in
    • 7archetypes and 5 behavioural indices
    • Privatestudents never see their own scores

    React · Vite · Supabase edge functions · Claude · jsPDF · Cloudflare

    1. Coach invites the student
    2. Student answers the intake
    3. Score: archetypes, indices, flags
    4. Claude writes a coaching insight
    5. Monthly check-ins are scheduled
    6. Trends and cohort analytics
    mta-psych Open full screen ↗

    How I build

    Every system starts with one real workflow. I build it fast with AI, test it, and keep a person in control of anything that goes out.

    1. Start from a real workflowI write down each step, input and decision of a team, a client or my own trading.
    2. Pick the simplest toolOften a Make.com, n8n or similar no-code flow is enough. When a workflow needs memory, rules or its own screens, I build an app.
    3. Build with Claude CodeI plan the system first. Then I build it in small steps that I review and test.
    4. Test every changeAutomated checks run before anything ships. Toodegrees OS alone has more than 115 test suites.
    5. Keep a person in controlAnything that publishes or sends needs an approval or a confidence threshold. Paid AI runs have a budget cap.
    6. Ship and measureI ship each system to the people who use it and track the results every week.

    Toolbox

    AI

    Claude CodeClaude APIOpenAIOllama (local models)RAG and embeddings

    Automation

    Make.comn8nWebhooks and APIsScheduled jobs

    Build

    TypeScriptReactNext.jsNode.jsPython

    Data and hosting

    SupabaseSQLiteCloudflareDockerVPS

    Marketing stack

    MailmodoWhopFigmaRemotionDataForSEORybbitSearch Console

    Experience

    Trading since 2020. Marketing and AI systems since 2026.

    2026 – now

    Marketing Operations & AI Systems

    Toodegrees · trading education and TradingView tools · remote

    • Run four marketing streams: email, YouTube, Discord and a monthly Intermarket Report. Content ships six days a week.
    • Write three weekly emails and segmented winback campaigns, with conversion tracking on every send.
    • Produce a weekly livestream, long-form videos, shorts and prop-firm reels.
    • Run weekly SEO and GEO for toodegrees.trade. Sessions per 28 days grew 36% and top-10 keywords grew from 1 to 5. See the case study.
    • Designed and built Toodegrees OS, the team's AI workspace.
    2023 – now

    Co-Owner

    Trading Pit Tribe · part-time · forex markets · remote

    2020 – now

    Forex Trader

    Self-employed · FX and futures, portfolio risk management

    2023 – 2024

    Junior Investments Specialist

    Saxo Bank · Amsterdam

    2021 – 2022

    Intern, then part-time employee

    L'Atelier Tailoring · Amsterdam

    2017 – 2019

    Sales Associate

    The Society Shop · Lelystad

    Markets & trading

    FX & futuresMacro analysisCOT positioningIntermarket researchRisk management

    Marketing

    Email lifecycleYouTube & shortsSEO & GEOAffiliate marketingDiscord communityTechnical writing

    AI & building

    Claude CodeMulti-agent workflowsMake.comn8nRAG & knowledge basesTypeScript / Next.js / ReactNode.jsSupabaseCloudflare

    Education

    Minor FinTech: Investing & Trading
    The Hague University of Applied Sciences · 2021–2022

    Fashion & Management, Marketing Innovation
    Amsterdam Fashion Institute · 2018–2022

    Market Structure Science
    New York Institute of Finance · 2023

    Financial Market Analysis
    International Monetary Fund · 2023

    SEO & GEO case study

    Three sites, one weekly method, from August to October 2026. Traffic numbers are 28-day windows from Rybbit analytics. Rankings come from DataForSEO and Google Search Console.

    1. BaselineRecord sessions, keywords and backlinks before any change.
    2. Fix blockersIndexing, canonicals, redirects and crawler access come first.
    3. Map AI answersAsk ChatGPT, Perplexity, Gemini and Claude the same buyer questions.
    4. Close gapsLog each gap. Ship the page or the structured data that fills it.
    5. MeasureCheck the numbers every week and rescan the AI engines every month.

    Toodegrees

    toodegrees.trade · TradingView indicators for ICT traders · US and 12 other markets

    Total sessions

    2,757

    Organic sessions

    612

    US ranked keywords

    31

    US keywords in the top 10

    5

    Referring domains

    52

    Pages crawled but not indexed

    59

    Visits from AI answers, per month

    Sessions that arrive from ChatGPT, Perplexity, Gemini or Claude, measured in Rybbit. Since January: ChatGPT 48, Perplexity 13, Gemini 5, Claude 3. AI visitors opened the pricing page 18 times and the checkout 4 times.

    The SEO work started in August. September was the best month on record. *October up to the 21st.

    What I did

    • Removed a robots.txt rule that blocked AI crawlers, and added a check that stops it from coming back.
    • Redirected /levelup. It sent 841 sessions a month to a 404 page.
    • Found why pages were not indexed and cut the list from about 140 to 59 (early July to 20 September).
    • Wrote six AI-search gap briefs and the explainer posts for them.

    Still open

    AI engines recommend Toodegrees on brand questions, in 8 of 36 engine checks. Non-brand questions are the next target.

    Mindful Trading Academy

    mindfultradingacademy.nl · trading psychology · NL and BE

    Google clicks (311 impressions)

    13

    Organic sessions, 28 days

    41

    Visits from AI answers, 28 days

    6

    4.6% of all MTA sessions came from ChatGPT answers. That is 13 times the share at Toodegrees.

    What I did

    • Made the site fully bilingual: 18 Dutch and English page pairs.
    • Wrote 7 pillar pages.
    • Fixed a soft-404: every broken URL showed the homepage. Added 6 redirects.
    • Added llms.txt and structured data for FAQs, terms and blog posts.

    Still open

    Rybbit stopped recording on 29 September, so October data is missing. The next step is a rescan of the four AI engines.

    Daphne Brands Coaching

    daphnebrandscoaching.com · online strength coaching for women · EN and NL

    Foundation phase · started 10 Aug 2026

    What I did

    • Fixed the booking path: 28 blog calls to action now open Calendly instead of an email link.
    • Built the Dutch /nl/ version of the site.
    • Rewrote 6 blog posts with FAQ structured data.
    • Added llms.txt and an allowlist for AI crawlers.
    • Fixed a conflict between the www and non-www versions of the site.

    Still open

    Traffic results are not in yet. The next step is the first content cluster.

    Contact

    edgarlopessalvador@gmail.com
    Download CV (PDF)

    Tilburg, the Netherlands · Dutch and English · The demos on this page run in your browser with sample data. The real systems use live data and need a login.