FLP Automation · Infrastructure Map

🔮

Boolean Builder · Proprietary Fintech Company Database

Foundation of search quality across P01 and P02 · built in-house · not available elsewhere

What it is

A curated database of 8 500 fintech companies mapped by country and category. Stored as companies_lookup.json and bundled directly in the Chrome extension, no external API call needed.

Coverage status

✓ EMEA: well stocked · high density across UK, France, Germany, Benelux & Nordics
→ Global: expanding across North America, APAC & LATAM

How it's used

P01: Daemon auto-classifies each JD by fintech category (Claude Haiku), loads up to 200 matching companies from the Boolean Builder, resolves Apollo.io org IDs in parallel (12 threads), and passes them as organisation filters to Apollo People Search, no extension config required.
P02: Source of the daily BD target company list: 1 company per day, picked by region + category.

Inbound JD Sourcing

Feed a job posting → auto-source LinkedIn → evidence-gated judgment → review HTML with labelling in 10 min

Trigger: JD URL (job board · ATS · Google Doc)

👤

You

Feed a JD

Job board, ATS, or Google Doc

1 click

🧩

Browser Extension

Extract & Build Criteria

Title variants · seniority · location · 50 companies · industry

instant

🗄️

Notion API

JD + 3 Criteria Queued

Broad · industry (BB companies) · direct competitors

≤10 min

⚙️

Python Daemon

Polls & Triggers

macOS launchd · always on

per criterion

🔍

Apollo.io API

Search People DB

Title + location + up to 200 fintech companies · 50 results · deduped by Apollo ID

per candidate

⚖️

Judgment Layer

Score & Bucket

Knockout pre-filter · evidence-gated Claude Haiku · interview / borderline / reject

review

📋

HTML File + Label

Review & Label

Auto-opens · all buckets shown · label → Golden Set → Notion write-back

Swim lanes

👤 You

Any browser

Open job posting → Click extension → Confirm preview → Review Notion shortlist (Fit Score ≥ 6)

🧩 Browser Extension

Chrome MV3 · JS

Extract page text + URL → Detect title / location / seniority → Pick 50 seed companies from Boolean Builder DB (8 500 cos · geo + category) · daemon expands to 200 → Generate 3 criteria JSON → Dedup vs Active-only → push to Notion

🗄️ Notion

Notion API v1

Store JD + full text → Store 3 criteria (Status: Active) → Log 1 Sourcing Run per JD → Mark criteria Done → Store candidates with Fit Score

⚙️ Python Daemon

Python 3 · launchd

Polls every 10 min → Fetch Active criteria from Notion → Classify JD category (Claude Haiku) → Load Boolean Builder · resolve up to 200 Apollo org IDs (12 threads) → POST to Apollo People Search API → Save candidates + CSV backup → Trigger Claude scoring

🔍 Apollo.io

People Search API · Basic plan

person_titles filter + person_locations filter + organization_ids (up to 200 fintech cos) + seniority filter (broad criterion) → 50 profiles per criterion · no credits consumed

✨ Claude Haiku

Anthropic API

Classify JD category → load BB company pool → Knockout pre-filter (deterministic Python): title/headline signals tech/ops role → hard reject, no LLM cost → Evidence-gated scorer: score ONLY on explicitly stated text: function · domain · seniority · lang/geo (0–10 each) → Bucket: interview (≥8 + med/high conf) · borderline (5–7) · reject (<5 or KO or insufficient data) → Write scores + evidence + bucket back to Notion Candidates DB · generate review HTML

📋 Review HTML

Auto-opens locally

All candidates shown: interview · borderline · reject · sorted by score → Per card: subscores · per-dimension evidence · reason · watch points · LinkedIn → Label buttons → save to Golden Set DB with reason

Business Development Outreach

Boolean Builder company → find relevant open role → source candidates → find HM → pitch with anonymised candidate

Trigger: Daily at 9am · 1 company per day (cron / launchd)

🔮

Boolean Builder

Pick Company

1/day from queue · filtered by region + category

scrape

✨

Claude Sonnet

Find Relevant Role

Job board scrape · filter: seniority · comp · relevancy (no pure IT)

source

🔍

Apollo.io

Source Candidates

Pipeline 01 on matched role · scored by Claude Haiku

find HM

🔍

Apollo.io

Find Hiring Manager

Search by title at target company · Head of TA / VP Sales / CRO

enrich

🔶

Clay

Enrich + Signals

Verify contact · funding · hires · personalisation data

draft

✨

Claude Sonnet

Pick, Anonymise & Draft

Best profile · strip identity · personalised email + LinkedIn message

send

📨

Lemlist

Multichannel Outreach

Email + LinkedIn connection + InMail · D+3 / D+7 follow-ups

log

🗂️

CRM

Log + Sequence

Contact + full multichannel sequence queued

Swim lanes

👤 You

One-time setup

Select region + category in Boolean Builder → seed BD queue → Review drafted emails before send (optional gate) → Monitor replies in CRM

🔮 Boolean Builder

Fintech DB · 8 500 cos

Filter by region + category → Pick 1 company/day from queue → Look up job board URL for company

✨ Claude Sonnet

Anthropic API · job filter

Scrape open roles at target company → Filter: exclude pure IT/eng · keep commercial · product · revenue roles → Score by seniority + earning potential + relevancy to FLP speciality → Select 1 best-fit role to pitch on

🔍 Apollo.io

People Search API · candidates

Run Pipeline 01 sourcing on selected role → Score + rank via Claude Haiku → Select top 1–2 shortlisted candidates

🔍 Apollo.io

People Search API · HM finder

Search by title at target company (org filter) → Target: Head of TA · VP Sales · CRO · COO → Reveal email (1 Apollo credit) → Return: name · title · email · LinkedIn URL

🔶 Clay

clay.com · enrichment

Verify email via waterfall (Apollo → Hunter → Dropcontact) → Confirm LinkedIn URL · scrape profile headline → Surface signals: recent funding round · headcount growth · open roles at company → Pass enriched data to Claude for personalised draft

✨ Claude Sonnet

Anthropic API · drafting

Anonymise candidate: "7 yrs KAM · Adyen + Worldline background" → Draft personalised BD email (HM name · role · candidate hook · Clay signal) → Write LinkedIn connection note (300 char) + InMail variant → Generate D+3 / D+7 follow-up variants (email + LinkedIn message)

📨 Lemlist

Email + LinkedIn sequences

Step 1, Email: personalised pitch with anonymised candidate → Step 2, LinkedIn: send connection request + personalised note to HM → Step 3, InMail: direct message on LinkedIn if 1st degree / no email reply → D+3 / D+7 follow-ups: email + LinkedIn message → Track: opens · clicks · LinkedIn accepts · replies → pause on positive signal

🗂️ CRM

HubSpot

Create contact (HM + company) → Log candidate bait reference + Clay signals → Sync Lemlist sequence status → Tag: BD Outreach · Pending Response

CV Sendout

Forward candidate email + CV → auto-label → Google Doc + Gmail draft → Fabien reviews + sends

Trigger: Forward to dedicated intake address · Gmail filter auto-applies FLP Process

👤

You

Forward to CV Sendout

CV attached · free-form notes · 1 click

instant

📧

Gmail Filter

Auto-apply FLP Process

Dedicated intake alias · zero clicks

auto

⚙️

run_pipeline.py

Orchestrate on Mac

Python 3 · macOS launchd

~5 sec

✨

Claude Haiku

Extract + Pitch

Fields · French pitch · 3–5 §

~3 sec

📄

Docs API

Copy Template + Fill

{{placeholders}} · pitch text · Drive folder

~20 sec

🖼️

pdftoppm + Drive

CV → Images

poppler · PNG · public URLs · inline in doc

~2 sec

✉️

Gmail API

Draft Reply in Thread

Doc link + PDF link · OAuth2

Swim lanes

👤 You

Gmail

Forward to CV Sendout intake address, any format, dump notes freely → Review Google Doc: adjust pitch · complete "À compléter" fields → Send Gmail draft

📧 Gmail Filter

Gmail · Rules

Match: dedicated intake alias → Apply label: FLP Process → Label removed by pipeline after processing, no reprocessing

⚙️ run_pipeline.py

Python 3 · macOS

Search Gmail for FLP Process label → Download email body + CV PDF attachment → Anchor: CV filename → subject → first body line for Name/Role/Client → Call Claude Haiku · create Drive folder · copy + fill Doc template → Convert CV → images → upload → insert inline · create Gmail draft

✨ Claude Haiku

Anthropic API

Read subject + CV filename + free-form body (any format, no structure required) → Extract: Name · Role · Client · Location · Nationality · Rate · Availability · Languages · LinkedIn → Generate French pitch: Background "Bref," · Fit "Ce qui colle pour X :" · Challengers · Pour info

📄 Google Docs API

Google API · OAuth2

Copy FLP template doc into Drive: Root / Client / Role → Replace {{NAME}} {{LOCATION}} {{POSITION}} {{RATE_SALARY}} {{AVAILABILITY}} {{LANGUAGES}} {{LINKEDIN_URL}} → Insert pitch text · append CV images inline

🖼️ pdftoppm + Drive

poppler · brew · Drive API

pdftoppm (brew install poppler): CV PDF → PNG per page, Mac-native → Upload each PNG to Drive · make public (anyoneWithLink) → insertInlineImage via Docs API · lh3.googleusercontent.com/d/{id}

✉️ Gmail API

Google API · OAuth2

Create draft reply in original thread → Body: link to Google Doc + link to original CV PDF → FLP Process label removed, thread won't reprocess

💡 No format required: dump call notes, LinkedIn text, forwarded emails, anything. Claude Haiku extracts all structured fields. Missing fields are filled with "À compléter" for Fabien to complete before sending.

Full Mandate Launch

Exploration call + JD → everything created automatically → sources until target headcount is reached

Trigger: Call transcript or notes + JD link

🎙️

Fathom / Otter

Record & Transcribe

Exploration call with client

extract

✨

Claude Sonnet

Parse Mandate

Role · seniority · location · headcount · skills · languages · comp · timeline

create

🗂️

CRM API

Build CRM Mandate

New deal · stages · contacts · timeline

generate

🔍

Claude Sonnet

Build Full Search

All search filters: title · co · industry · seniority · location · skills · language

generate

✉️

Claude Sonnet

Write Outreach

1st touch + 2 follow-ups + LinkedIn note

run

⚙️

Pipeline 01

Auto-Source

Score all candidates via Haiku

check

🔁

Python Loop

Loop Until Target

If shortlist < 40/headcount → expand criteria → re-run

done

✅

You

Notify & Review

Mac notification + Notion view ready

Swim lanes

👤 You

Fathom + Cowork

Run exploration call (Fathom records) → Paste transcript + JD link into Cowork → Approve mandate brief before launch → Review 40-candidate shortlist when ready

🎙️ Fathom

Fathom / Otter.ai

Auto-join call · transcribe in real time → Export transcript + summary

✨ Claude Sonnet

Anthropic API

Extract structured mandate: role · seniority · headcount · location · comp · timeline · ideal companies → Build full search JSON (all 7+ filters) → Generate outreach sequences (1st touch + D+3 + D+7 + LinkedIn) → Expand criteria if loop needs more (relax geo · add title variants)

🗂️ CRM

HubSpot / Vincere / Notion

Create mandate deal (client · role · stage: Intake) → Set pipeline stages (Intake → Sourcing → Shortlist → Interview → Offer) → Link contacts (client-side HM + internal) → Log all sourcing activity + cost

⚙️ Pipeline 01

Apollo + Claude Haiku

Run structured Apollo People Search (all criteria) → Score each profile vs mandate brief → Dedup across all runs by LinkedIn URL

🔁 Python Loop

Sourcing daemon

Count candidates with Fit Score ≥ 6 → If < 40 × headcount → ask Claude to expand criteria → Re-run sourcing (max 5 iterations) → Notify + stop when target reached

📨 Lemlist

Email + LinkedIn sequences

Import shortlist candidates with LinkedIn URLs → Enroll in Claude-written multichannel sequence (email + LinkedIn connection + InMail) → Track: email opens · LinkedIn accepts · replies → Pause sequence on positive signal · sync reply to CRM

Candidate Outbound Placement

Chrome extension intake → parse CV → scrape jobs → find HM → score + draft 3 pitch variants → HTML review on Desktop → Lemlist launch

Trigger: Chrome extension on linkedin.com/in/* page

🧩

Chrome Extension

Intake

LinkedIn URL · CV PDF · LinkedIn PDF · location · sectors · company sizes · language pref · notes

parse

⚙️

pdfminer.six

Parse CV

CV + LinkedIn PDF → experience / education / skills

assess

✨

Claude Sonnet

Assess Profile

Summary · title variants · strengths · inconsistency flags

scrape

🔎

X-ray + Job Boards

Scrape Jobs

20+ ATS (Greenhouse/Lever/Ashby…) + WTTJ/Cadremploi/APEC/Indeed FR · 14-day filter

find HM

🔍

Apollo.io

Find Contacts

2 HM + 2 TA per company · capped at 30 companies/run

score

✨

Claude Haiku

Score Targets

1–10 relevancy per target vs candidate profile

draft

✨

Claude Sonnet

Draft 3 Variants

Anonymised pitch · no candidate name · no employer names · under 120 words each

review

📋

HTML File

Review on Desktop

Drops on Desktop · opens automatically · Skip / Manual / Approve per card

launch

📨

Lemlist + LinkedIn

Launch Outreach

Email D0/D7/D14 · LinkedIn queue via extension · GCal daily to-do

Swim lanes

👤 You

Chrome Extension

Open candidate LinkedIn profile → click extension → Upload CV PDF + LinkedIn PDF · add notes · pick sectors/sizes · set language → Review HTML file on Desktop: Skip / Manual / Approve per card → Click Launch → Lemlist sequences start + GCal to-do updated

🧩 Chrome Extension

Chrome MV3

Extract LinkedIn profile data (name · title · location) → Encode CV + LinkedIn PDF as base64 → Send intake to local daemon → runs in background → Poll LinkedIn queue every 5 min · draft LinkedIn messages (user hits Send manually)

⚙️ Python Daemon

Flask · macOS launchd

pdfminer.six: extract CV text + LinkedIn PDF text → Heuristic parser: split into experience / education / skills → X-ray DuckDuckGo: 20+ ATS (Greenhouse, Lever, Ashby, Teamtailor...) + WTTJ/Cadremploi/APEC/Indeed FR · 14-day freshness filter → Funding signals via DuckDuckGo HTML scrape · basic email verification → Build HTML review file → drop on Desktop → open automatically

✨ Claude Sonnet

Anthropic API · assessment

Assess profile: summary · title variations · strengths · CV/LinkedIn inconsistency flags · candidate language detection → Draft 3 anonymised pitch variants per target (no candidate name, no employer names, under 120 words)

✨ Claude Haiku

Anthropic API · scoring

3-stage judgment: Python knockout pre-filter (function · seniority · geo · language) → Haiku scores function / domain / growth / practical fit → Rule engine applies rules built from past skip reasons → Rationale per score · low scorers deprioritised in review

🔍 Apollo.io

HM + TA finder per company

Up to 2 HM + 2 TA per company · /mixed_people/search → Capped at 30 companies per run → Return: name · title · email · LinkedIn URL

🗄️ Notion

Notion API · backend only

Log run (status · candidate · timestamp) → Log each outreach decision (approved / skipped / manual + reason) → Queue LinkedIn messages (status: pending) · mark done after extension sends → Dedup: check company history across candidates before outreach

📨 Lemlist

Email sequences · D0/D7/D14

D0, Email: anonymised pitch via {{icebreaker}} variable · FR or EN campaign → D7, Follow-up bump → D14, Close-out email → Dedicated sending domain · replies forwarded to main inbox

💡 Under the hood: judgment system v2 (knockout + scoring + rules learned from past skip reasons) · unified Chrome extension · HTML review UI (per-recipient checkboxes, 3-variant carousel, FR/EN auto-detection) · email templates D0/D7/D14 in FR + EN. No Clay dependency: funding signals via web scraping.

🧮 Cost & Output per Run

One run = one full pipeline loop. Conservative estimates. Flat subscriptions amortised separately.

	P01 · per JD	P02 · per company	P03 · per CV	P04 · per mandate	P05 · per candidate
💰 Cost per run conservative estimate	~$0.06	~$0.60	~$0.01	~$0.70	~$0.60
Expected output per run	~150 candidates sourced · scored & bucketed · review-ready shortlist in ~10 min	1 target company · 1 best-fit role · 1–2 anonymised candidates · 1 HM in multichannel sequence	1 client-ready CV dossier (Google Doc) + Gmail draft in under a minute	Full mandate: CRM deal + search + sequences · 40+ qualified candidates per headcount	Up to 30 target companies · up to 120 HM/TA contacts · 3 pitch variants each · D0/D7/D14 sequences
Main tools	Apollo · Claude Haiku · Notion · Boolean Builder	Boolean Builder · Claude Sonnet · Apollo · Clay · Lemlist · HubSpot	Gmail · Claude Haiku · Google Docs/Drive	Fathom · Claude Sonnet · Apollo · Lemlist · CRM	Apollo · Claude Sonnet + Haiku · Notion · Lemlist
Subscriptions used flat monthly	Apollo	Apollo · Clay · Lemlist	None · free stack	Lemlist	Apollo · Lemlist

🎯 FLP Automation · Infrastructure Map

Inbound JD Sourcing

Business Development Outreach

CV Sendout

Full Mandate Launch

Candidate Outbound Placement

🧮 Cost & Output per Run