Anuj Kapoor Get in touch

For Alex & Dan · Application - Deel Ghostbuster · 2026-05-10

Ghostbuster work is what I do for a living.
I just don't have the title yet.

Built 40+ production agents across 8 functions at Finmo (an Airwallex-class APAC competitor) in 21 days, mostly solo, on Claude Code. Same operating thesis as Deel - cross-border, multi-entity, compliance-heavy, high-touch onboarding. Same ghosts. I have already done the boring parts.

Anuj Kapoor
Anuj Kapoor Director, BD / Pre-sales / Onboarding · Finmo Singapore · APAC
40+
Agents shipped
8
Functions automated
21
Days, mostly solo
60+
Hours saved / week
See the receipts → anuj@scale-with-clarity.com Singapore · Remote-ready

The receipts

Alex's words on the left. My evidence on the right.

Direct mapping from the LinkedIn post to systems already in production. The catalogue underneath proves each line.

From the post
My evidence
"A process installed by someone who left 2 years ago."
Killed a $10K/month outsourced SDR contract at Finmo. Replaced it with an in-house outbound engine I built and ran myself before hiring around it. The whole replacement infra runs at under $500/month.
"A tool slowing everyone down that nobody wants to call out."
SDRs were tab-switching between multiple dialer providers depending on which had better connectivity for a given country. I integrated the APIs of all of them into a single in-house Next.js + Twilio WebRTC + Exotel PSTN dialer. One window, smart provider routing, every leg auto-logged to HubSpot. 601 dials in two weeks, 32.8% pickup, zero manual switching.
"A decision from 2022 that no one revisited."
DLSP (16-school PH education network) was about to release a treasury RFI. First multi-entity education-network treasury deal Finmo had ever attempted - zero prior playbook. I refused to respond to the RFI. Reframed the sales pitch into a joint discovery: an 8-week TOM Design Sprint where we co-authored the RFI together. Won the architecture before anyone else wrote a proposal.
"Ghostbusters find broken things and fix them. Own the outcome end to end."
120+ merchant onboardings handled solo in year one at Finmo, SMB to enterprise, APAC to EU. Today running BD + Pre-sales + Onboarding as a single function from sourced to live.
"Builder mentality. You'd crush it without resources."
First Claude Code session: 2026-04-02. 40+ agents, 8 functions, 6 production dashboards, this page: 21 days later. Solo, on a working laptop, while running the day job.
"First-principles thinker. 'Why does this even exist?'"
Reframed an EastWest Bank treasury deal from a TMS replacement (which would have triggered margin-conflict resistance) to an "invisible command layer" over four heterogeneous cores. Bank keeps 100% of FX P&L. We sit on top.
"Financially sharp. Business-minded. Operationally elite."
Negotiated and signed off four hard success metrics before kickoff at DLSP: consolidated position <10 min (vs ~2 days), >=85% auto-match on first-pass recon, <15% forecast variance at week-4, zero cross-school data leak.
"AI-native. Help every team rethink how they work, every single week."
Built a Retell + Twilio + Claude voice AI in 3 days. 191 calls, 73% live contact, $13.50 total spend. Wrote the verdict honestly: channel works, pitch needs redesign - and learned more in 3 days than 3 months of vendor demos.
"Self-directed. Mission, not playbook."
Joined Finmo as a founding-team member when the company was 20 people. Pre-sales didn't exist. Founders were doing it between product sprints. Took it off their plate. Built the team. Promoted to Director.
"Partner directly with Dan and myself on highest-priority initiatives."
Most comfortable working directly with founders and CXOs. Did it at Finmo from day one (under-20 company, founders in the same Slack). Did it at Mascloft cold-calling CEOs at D&B, HP, Taj Group, GroupM. Did it at DLSP and EastWest at sponsor level.

Function-by-function catalogue

40+ ghosts taken down across 8 functions.

For each one: what it decides or automates, the stack, and the volume / outcome metric. Each one replaced a manual process, a SaaS, or a recurring meeting.

SALES / SDR

Outbound runs itself

Lead enrichment to call dispositioning to next-best-action - every step idempotent, every step measured.

Daily SDR Task Generator

Ensures every lead gets followed up properly. Zero lead leakage, zero delayed follow-ups, and SDRs never have to create a single task manually. The system picks the 50 best calls / LinkedIn / WhatsApp tasks per SDR per day using intent score, recency, and channel.

Stack
Python · HubSpot API · Airtable · GitHub Actions
Outcome
70% completion · 0 unattended leads in 4 months · 50 tasks/SDR/day auto-generated

SmartReach Campaign Orchestrator

Enabled dynamic, per-prospect content at campaign send time. SDRs no longer have to manually craft and send custom emails to each prospect. Pushes prospects, assigns campaigns, syncs reply sentiment back to HubSpot.

Stack
Python · SmartReach v3 · HubSpot
Outcome
4,089 prospects in active campaigns · 1,839 emails sent in 2 weeks · 64% open rate

AI SDR Cold Caller (Retell)

Did full ICP validation and market research in 3 days end-to-end. Runs ACA loop + BANT qualification + voicemail detection + warm transfer on outbound calls. The pilot answered "is this segment reachable" with a real number on a real budget, instead of a quarter-long debate.

Stack
Retell AI · Twilio SIP · Whisper · Finmo product KB
Outcome
191 calls in 3-day pilot · 73% live contact · $13.50 spend · ICP validated, dead segments killed

In-house Sales Dialer

SDRs were tab-switching between multiple dialer providers because each had better connectivity in different countries. I integrated all of their APIs into one in-house Next.js dialer. One window, smart provider routing, every call leg auto-logged to HubSpot. No more manual switching.

Stack
Next.js · SQLite · Twilio WebRTC · Exotel PSTN · HubSpot
Outcome
601 dials in 2 weeks · 32.8% pickup · zero manual switching · zero manual logging

Outbound Call Sync

Pulls call dispositions every cron tick. Refreshes the SDR task queue based on outcomes.

Stack
Python · HubSpot Calls API (batch) · Airtable · cron
Outcome
317 leads synced weekly · status updates in minutes, not days

Slack SDR Briefing Bot

Three daily briefings (morning / midday / evening) with task counts, overdues, next-day forecast.

Stack
Python · HubSpot · Slack webhooks
Outcome
Killed manual stand-up overhead · SDRs land at desk pre-planned

Prospect Research Orchestrator

Spins up 4 parallel research agents per company (firmographics, signals, decision-makers, competitor map) and merges output.

Stack
Claude Code subagents · Apollo · web research
Outcome
4 hrs → ~20 min per account · used on 192 HIGH-priority leads
MARKETING / LEAD GEN

Source, score, enrich

Source-of-truth lists rebuilt nightly with a waterfall enrichment stack. No vendor lock-in.

LinkedIn Sales Nav Scraper

Persistent Playwright session, named-lead extraction from saved searches, hand-off to enrichment.

Stack
Python · Playwright · persistent browser context
Outcome
Replaces Phantombuster spend · 16,950 lines of session work

Lead Gen 3.0 (AU/SG)

Per-market target lists, dedupe against pipeline, ICP-fit tagging.

Stack
Python · Apollo · web research
Outcome
192 HIGH-priority AU leads · $13.7M projected pipeline

ICP Scorer

Scores each lead 0-100 against industry, headcount, geography, revenue band, tech stack.

Stack
Python · Claude scoring · Airtable
Outcome
Drops bottom quartile before any enrichment spend

Lookalike Builder v3

Reverse-engineers competitor customer lists (Atlar, HighRadius, Palm), finds APAC lookalikes.

Stack
Python · Apollo org enrichment · pattern matching
Outcome
30 named competitor customers analysed · 50+ lookalikes surfaced

Lead Enrichment Tool (one URL, full waterfall)

Sales team was juggling 5 separate platforms - ContactOut, Apollo, Hunter, Clearout, plus manual LinkedIn. Replaced all of it with one tool. Hits cheapest provider first, falls through on miss, stops on verified hit. 4 minutes → 8 seconds per lead.

Stack
Python · Prospeo · BetterContact · Apollo · ContactOut · Clearout
Outcome
334 leads enriched · ~60% lower cost-per-verified-contact than any single vendor · SDRs stopped tab-switching
CUSTOMER SUCCESS / ONBOARDING

Solutioning leaves the slide deck

Discovery, solutioning, and onboarding live in dashboards instead of decks. One source of truth per merchant.

Solutioning Engagement Runbook

Standard playbook for free pre-POC discovery sprints. Defines gates, success metrics, stakeholder map.

Stack
Documented runbook · used to feed solutioning dashboards
Outcome
Landed a 16-entity education network and a 5-entity listed bank · both moved cold to signed pilots

DLSP Solutioning Dashboard

Multi-entity onboarding state, gates, sandbox / UAT / go-live status, escalations.

Stack
Next.js · Supabase · Lovable
Outcome
Replaces a 30-page RFI doc with a live dashboard · 16 schools tracked

EastWest Solutioning Dashboard

Same playbook, applied to a 5-entity listed universal bank with 4 heterogeneous core systems.

Stack
Next.js · Lovable
Outcome
Discovery script + solutioning approach + dashboard, generated end-to-end

Veem KYB Onboarding Exporter

One-line Python tool. Pulls 5 native SQL queries against the org DB, assembles a Veem-shaped 50-row CSV (33 business + 17 BOI per shareholder ≥25%).

Stack
Python · Metabase · MySQL
Outcome
Killed a recurring manual export tax · scaled onboarding throughput

Merchant Lifecycle View

Single dashboard joins sign-up to KYB to onboarding to UAT to go-live to revenue. Per merchant, per partner.

Stack
Metabase MBQL · HubSpot pipeline · Jira Ops · partner-hierarchy joins
Outcome
Replaces 4 separate weekly reports · leadership stops asking "where is X merchant"
RISK / COMPLIANCE

Decisions, not just rules

KYB triage agent decides which sign-ups deserve a human. Junk gets parked, real merchants get routed.

KYB End-to-End Pipeline (webhook → HubSpot)

Sign-up webhook → Zapier intake → Python classifier (100+ patterns: legit / junk / employee-test) → Apollo org enrichment + dedupe → round-robin to HubSpot contact + deal → Slack ping to the assigned SDR. The whole pipeline lives without anyone touching it.

Stack
Zapier · Python · Apollo · Airtable · HubSpot · Slack
Outcome
47% of sign-ups auto-rejected · 100% legit-to-CRM match · 0 unattended sign-ups in 4 months · ~10 hrs/wk saved

Email Verification Pipeline

Batch validates outbound emails, drops disposable / free / gibberish before send.

Stack
Python · Clearout API
Outcome
Validates 334+ outbound leads · protects sender reputation

Phone Validation

Validates format, country code, carrier type before any dial attempt.

Stack
Python · Twilio Lookup
Outcome
Underpins 652+ dialed calls · prevents wasted credits
OPERATIONS

Daily ops pre-staged

Weekly reviews self-generate. CRM audit running continuously in the background.

HubSpot Portal Audit

Scans 521 contact properties, finds duplicates, flags stale custom fields, scores workflow health.

Stack
Python · HubSpot Properties + Workflows API
Outcome
12 duplicate field pairs found · 107 stale custom fields flagged

Ops Lead Enrichment v2

Catches every inbound, enriches the org, writes back to HubSpot, routes by region.

Stack
Python · HubSpot · Apollo · regex routing
Outcome
0 unattended leads in 4 months · 100% inbound match-rate to CRM

Weekly Review DOCX Generator

Pulls the week's data and renders a leadership review without manual cut-paste.

Stack
Python · python-docx · HubSpot · SmartReach · Metabase
Outcome
~3 hrs/week saved · Solutions Weekly + SDR Weekly in one run

Daily SDR Runbook (the system as documentation)

Cron-driven morning task generation, status sync, Slack notification chain.

Stack
GitHub Actions · Python · HubSpot · Airtable · Slack
Outcome
Onboards a new SDR or ops hire in one read
SUPPORT / INBOUND

Inbound is a conveyor, not a queue

Webhook triggers enrichment and routing in seconds.

Mixpanel intent tracking (built without engineering)

Marketing wanted to know when product-side intent fired ("Clicked Add Bank", "Started KYB") so they could trigger campaigns. Engineering backlog said "next quarter." I shipped the whole pipeline in a weekend. Mixpanel webhook → Apollo enrichment → HubSpot timeline → SmartReach campaign trigger → Slack ping the SDR.

Stack
Mixpanel webhooks · Zapier · Python · Apollo · HubSpot · SmartReach · Slack
Outcome
0 eng hours · marketing unblocked · lead-response latency hours → minutes

Inbound Audit (3-month)

Reconstructs every inbound from Zapier history, joins to HubSpot deals, flags unattended.

Stack
Python · JSON analysis · HubSpot
Outcome
422 sign-ups audited (Dec → Mar) · 233 real, 40 engaged, 0 unattended

Call Recording Analyzer

Downloads recordings, transcribes via Whisper, classifies tone (neutral / cold / hostile) and outcome.

Stack
Python · Whisper API · Exotel/Twilio recording APIs
Outcome
181 calls transcribed in 2 weeks · 78% neutral, 21% cold, 1% hostile
ANALYTICS / BI · LEADERSHIP

Leadership reads the same dashboard the team works from

No more weekly cut-and-paste from 6 systems.

MBR Dashboard (Monthly Business Review)

Pipeline value, SDR activity, deal progress, channel breakdowns, KPI trends. The deck for senior leadership - live.

Stack
React 19 · TypeScript · Recharts · Vite · Lovable · Supabase
Outcome
$4.71M total pipeline view · 23 deals tracked · 648 unique companies

Real-time Unified Dashboard

Pulls from HubSpot + Jira + Metabase + Twilio + Exotel + SmartReach into a single view.

Stack
React · Lovable · Supabase · multi-API aggregation
Outcome
6 source systems, 1 view · ~11 hrs/week saved

Solutions Weekly Dashboard

Per-week view for the Solutions org: active deals, blocked items, lifecycle stage.

Stack
Lovable · Supabase · HubSpot · Jira
Outcome
Replaces a 3-hour Friday compile · leadership reads it themselves

SDR Weekly Dashboard

Daily call volume, pickup rate, conversation quality, meetings booked, per SDR.

Stack
React 19 · TypeScript · Recharts · Lovable
Outcome
Tracks 2 SDRs across AU/SG · 274 + 313 dials in last cycle

Metabase Data Pull

Templated MBQL queries for revenue, invoice, and transaction views by org_id. Partner-hierarchy roll-ups.

Stack
Metabase API · MBQL · Python
Outcome
559 transaction columns mapped · daily aggregated txn counts/volumes/fees per org

Finmo Pulse

Real-time pulse view across the company. Designed for leadership glance-checks.

Stack
Lovable · React · Supabase
Outcome
Replaces ad-hoc Slack pings asking "how is the week going"
HR / TALENT (INTERNAL ATS)

Hiring managers get the same agent stack the SDRs do

Source, score, and prep without leaving the tool.

Candidate Sourcing Pipeline

LinkedIn search → enrichment → scoring → shortlist tracker. Same waterfall enrichment Sales uses.

Stack
Playwright · Apollo · Hunter · Airtable
Outcome
Reusable for any role · hiring manager runs it themselves

Candidate-JD Fit Scorer

Scores each candidate's CARs + resume against the role's JD. Surfaces fit, gaps, and risk.

Stack
Claude scoring · structured prompt · Airtable
Outcome
Initial screen: 30 min → 5 min per candidate

Resume Tailoring Engine

Ingests CAR statements + JD, generates a tailored LaTeX/PDF resume per role.

Stack
Python · LaTeX · Claude
Outcome
Removes the "rewrite my resume" bottleneck for internal mobility

Interview Prep Brief Generator

Compiles company background, recent hires, leadership, growth signals into a prep doc.

Stack
Web research · structured Claude prompt
Outcome
Shared with candidates before final-round · improves prep + signal

Featured systems

Six marquee builds. End-to-end, not point automations.

Each one makes decisions, not just executes tasks. Each one replaces a SaaS, a process, or a meeting.

In-house Sales Dialer - one window, multi-provider routing

SDRs were tab-switching between multiple dialer providers because each had better connectivity in different countries. I integrated all of their APIs into one in-house Next.js dialer with smart provider routing. One window, every leg auto-logged to HubSpot, no more manual switching.

Architecture
  • Browser ⇒ Twilio WebRTC token (in-app calling)
  • Country-aware router ⇒ pick best PSTN provider per dial
  • Outbound ⇒ Twilio + Exotel + alternates, single UI
  • Every call leg ⇒ HubSpot Calls API (batch sync)
  • Campaign manager ⇒ SQLite (next-best-call queue)
Stack
Next.js · SQLite · Twilio WebRTC · Exotel PSTN · HubSpot
Metric
601 dials in 2 weeks · 32.8% pickup · zero manual switching · zero manual logging

AI SDR Cold Caller (Retell AI)

Voice agent runs the same playbook as the human SDRs. Beats human pickup. $0.07/call.

Architecture
  • Retell AI agent ⇒ Twilio SIP trunk
  • Greeting ⇒ time check ⇒ intent ⇒ BANT ⇒ pitch ⇒ close
  • ACA loop on objections (Acknowledge / Compliment / Ask)
  • Voicemail detection + warm transfer to human SDR
  • Transcripts ⇒ Whisper ⇒ HubSpot activity log
Stack
Retell AI · Twilio SIP · Whisper · Finmo product KB
Metric
191 calls in 3-day pilot · 73% live contact · $13.50 spend

Unified Real-time Dashboard

Six source systems collapsed into one view. Leadership stops asking, the team stops cut-pasting.

Architecture
  • HubSpot (pipeline) + Jira (ops) + Metabase (revenue)
  • Twilio + Exotel (call activity) + SmartReach (email/LinkedIn)
  • Aggregator ⇒ Supabase ⇒ React/Recharts dashboard
  • Auto-refresh on schedule, no manual intervention
Stack
React · Lovable · Supabase · 6 source APIs
Metric
Replaces ~6 separate weekly reports · ~11 hrs/week saved

KYB Pipeline - webhook to HubSpot, end-to-end

Most sign-ups are noise. Every legit one is a customer waiting. The pipeline decides who deserves a human, enriches them, and routes them - all before anyone on the team sees the alert.

Architecture
  • Sign-up webhook ⇒ Zapier intake
  • Python Code step ⇒ classify (legit / junk / employee-test) via 100+ patterns
  • Junk ⇒ park in Airtable, never disturb humans
  • Legit ⇒ Apollo org enrichment + domain dedupe
  • Round-robin assignment ⇒ HubSpot contact + deal created
  • Slack ping to assigned SDR with full context
Stack
Zapier · Python · Apollo · Airtable · HubSpot · Slack
Metric
47% of April sign-ups auto-rejected · 100% legit-to-CRM match · 0 unattended sign-ups in 4 months · ~10 hrs/wk saved

Lead Enrichment Tool - one URL, full waterfall

Sales team was juggling ContactOut, Apollo, Hunter, Clearout, plus manual LinkedIn lookups. I built one tool with one URL. Hits the cheapest provider first, falls through on miss, stops on verified hit. SDRs went from 4 minutes per lead to 8 seconds.

Architecture
  • Single endpoint ⇒ Prospeo (cheapest, fastest)
  • Miss ⇒ BetterContact ⇒ Apollo ⇒ ContactOut waterfall
  • Verified hit ⇒ Clearout email-validation gate
  • Pass ⇒ HubSpot + Airtable write-back, source-attributed
  • Spend tracker ⇒ per-provider cost-per-verified-contact
Stack
Python · Prospeo · BetterContact · Apollo · ContactOut · Clearout · HubSpot · Airtable
Metric
4 min → 8 sec per lead · 334 leads enriched · ~60% lower cost-per-verified-contact than any single vendor

Mixpanel intent tracking - built without engineering

Marketing wanted to know when product-side intent fired - "Clicked Add Bank", "Started KYB", "Connected Bank" - so they could trigger campaigns. The engineering backlog said "next quarter." I shipped the whole pipeline in a weekend. No eng ticket. No product spec. No headcount ask.

Architecture
  • Mixpanel intent events ⇒ outbound webhook
  • Zapier intake ⇒ Python Code step
  • Apollo enrichment by email ⇒ org context
  • Dedupe against HubSpot contacts
  • Push event + org context ⇒ HubSpot timeline
  • Tag contact ⇒ trigger SmartReach campaign
  • Slack ping the SDR who owns the account
Stack
Mixpanel webhooks · Zapier · Python · Apollo · HubSpot · SmartReach · Slack
Metric
0 engineering hours · marketing team unblocked · lead-response latency hours → minutes · ~1,077 lines of session work

Impact

Time saved. Cost avoided. Decisions sped up.

Hour estimates are conservative, sourced from runbooks and SDR baselines. Volume numbers are pulled directly from the code.

WorkloadBeforeAfterSaved
Weekly leadership reports (HubSpot, Metabase, SmartReach, Twilio, Exotel)~12 hrs/wkAuto refresh~11 hrs/wk
Inbound lead enrichment + routing~3 hrs/dayWebhook (real-time)~15 hrs/wk
KYB sign-up triage~2 hrs/dayFilter agent (auto)~10 hrs/wk
Outbound dialing + call logging (per SDR)Manual log to HubSpotTwilio + Exotel + auto-log~5 hrs/wk per SDR
Lead enrichment (per SDR)4 min, 5 tabs8 sec, one URL (waterfall tool)~6 hrs/wk per SDR
Mixpanel intent → marketing campaignsEngineering backlog (next quarter)Webhook (real-time, weekend build)0 eng hours blocked
Outbound content authoring (per campaign)~2 days~30 min (gen + review loop)~15 hrs/campaign
Prospect deep-research (per account)~4 hrs~20 min (orchestrator)~3.5 hrs/account
Weekly review document~3 hrs (manual)Auto-generated DOCX~3 hrs/wk
Total recurring savings - - 60+ hrs/wk across team

Volumes processed (real numbers from the code)

Lead volume
  • 334 outbound leads end-to-end (196 AU + 138 SG)
  • 648 unique companies worked across all channels
  • 192 HIGH-priority AU leads · $13.7M projected pipeline
  • 422 sign-ups audited Dec 2025 → Apr 2026 · 0 unattended
Call volume
  • 652 calls (Feb-Mar) + 601 calls (Apr 2-15)
  • 32.8% pickup rate across both periods
  • 181 call recordings auto-transcribed via Whisper
  • 76 live conversations in 3-day Retell pilot · $13.50 spend
Email / outreach
  • 4,089 prospects in active SmartReach campaigns
  • 1,839 emails sent Apr 1-15 · 64% open rate
  • 853 contacts reached in single 2-day push
Pipeline
  • $4.71M total pipeline (base, Feb-Mar)
  • 23 deals in motion · 1 Closed Won · 1 Qualified to Buy at $813K
  • 100% match rate on legit inbound to HubSpot
  • 15 meetings booked from April inbound
Infrastructure
  • 120 Python scripts in repo root
  • 160+ markdown runbooks and analyses
  • 6 production dashboards (MBR, Solutions, SDR, Pulse, Real-time, Dialer)
  • 92 Claude Code sessions in 21 days · 4.4/day avg
  • 521 HubSpot contact properties audited (149 custom)
Cost avoided
  • SaaS sales dialer (replaced by in-house build)
  • External research / list-building agency spend
  • Additional SDR ops headcount (postponed)
  • Single-vendor enrichment lock-in (waterfall lowers cost-per-verified-contact)
  • $10K/month outsourced SDR contract (killed)

Side build

Ladder - same playbook, applied to my own job hunt.

When the APAC fintech market got brutal, I rebuilt my own job search the same way I'd rebuild a sales ops org. Multi-agent discovery, freshness verification, CAR-to-JD matching, and a personal CRM. Currently compresses what used to be ~39 hrs/week of manual searching down to ~5–8 hrs of human judgment.

What it does

A single-user platform - branded Ladder, on the way to ladderhq.ai - for layoff and career-break candidates in India and APAC. Inspired by nxtjob.ai, but discovery-first instead of resume-first. Wraps an 8-method search strategy around a verification moat (76% of fresh listings on a random day in April were already expired - that's the gap competitors haven't closed).

  1. Profile - collect titles, locations, seniority, industry, salary floor, freshness window.
  2. Parallel discovery × 3 agents - careers pages (Agent A), job portals (Agent B), creative search / LinkedIn hiring posts / Google dorks (Agent C).
  3. Consolidation + dedupe - across discovery agents into one ranked list.
  4. Parallel freshness verification × 3 - every URL re-checked still-live before it surfaces to the user.
  5. CAR ↔ JD scoring - every JD scored against my 80-statement CAR bank to surface only high-fit roles.
  6. People discovery - hiring manager, talent partner, 1st/2nd-degree referrers per job, via Unipile + Prospeo + BetterContact waterfall.
  7. Resume + cover letter - LaTeX templates compiled per role in GitHub Actions, customised by the matched CARs.
  8. Outreach + CRM - Airtable as the personal CRM with status tracking and reply sentiment back-sync.

Public artifacts

Code, dashboards, and demos you can click through.

Some of the work is private (Finmo internal). These are the public bits - repos and Lovable apps you can open in a new tab and poke at.

GitHub repos
Lovable apps

Stack

~30 APIs catalogued. Most of them in production.

Comfortable reading docs, running Postman, diagnosing webhook + auth failures without engineering support.

LLMs
  • Claude (Code, Sonnet, Haiku, Opus)
  • GPT
  • Whisper (transcription)
Orchestration
  • Claude Code (subagents, hooks)
  • Zapier
  • n8n
  • GitHub Actions
  • Google Apps Script
  • cron
Voice
  • Twilio (WebRTC + Voice + Lookup)
  • Exotel (PSTN APAC)
  • Retell AI
CRM / Data
  • HubSpot
  • Airtable
  • Metabase
  • Supabase
  • SQLite · PostgreSQL
Comms
  • SmartReach (v3 API)
  • Unipile
  • Phantombuster
  • Gmail API
  • Slack webhooks
Enrichment
  • Apollo
  • Prospeo
  • BetterContact
  • Clearout
  • ContactOut
  • People Data Labs
Frontend
  • Next.js / React 19
  • TypeScript
  • Tailwind
  • Recharts
  • Vite · Lovable
Ops / Misc
  • Jira (Ops board)
  • Confluence
  • Mixpanel
  • Freshworks

Why this matters for Deel

Same thesis. Same compliance load. Already automated.

Same operating thesis

Cross-border, multi-entity, compliance-heavy, high-touch onboarding. Same boring parts to automate. I have done it in your category - at Finmo (treasury / payments), Exotel (CPaaS), and earlier in HR-tech. The shape of the ghost is the same, only the rails change.

Speed proof, not promise

First Claude Code session: 2026-04-02. 40+ agents, 8 functions, 6 production dashboards, this page: 2026-04-23. 21 days. Solo. While running BD + pre-sales + onboarding as the day job. The Ghostbuster role is asking for this exact tempo.

Scope bias toward outcomes

The Ghostbuster JD says "own the outcome end-to-end." That's literally what I do at Finmo today - sourced to live, including the implementation. I don't hand off to an engineer or a CSM and lose the thread. The whole point is staying with the deal until the customer is using it.

Competitive intel built in

I know how a smaller competitor uses AI to compress ops cost. That's useful from day one inside Deel - both for benchmarking and for spotting which incumbent processes are actually replaceable now versus next year.

Founder-direct comfort

Most comfortable working directly with founders and CXOs. Did it from day one at Finmo (under-20 company). Did it at Mascloft cold-calling CEOs. Did it at DLSP and EastWest at executive-sponsor level. "Partner directly with Dan and myself" is how I work, not a stretch.

I write the runbook while running it

Every system above has a markdown runbook in the repo. The system is the documentation. That matters when a Ghostbuster fix has to outlive the Ghostbuster.

Career

Eight years across cross-border payments, treasury, and CPaaS.

  1. Jan 2025 - present

    Director, BD / Pre-sales / Onboarding · Finmo

    Singapore. Lead BD, pre-sales, and onboarding as one team that owns the full deal-to-go-live journey. Built the in-house outbound engine that replaced a $10K/month outsourced setup.

  2. Nov 2023 - Jan 2025

    Manager, Customer Enablement & Success · Finmo

    Bengaluru. Founding-team member when the company was under 20 people. Built pre-sales and onboarding from zero. 120+ merchant onboardings solo in year one.

  3. May 2021 - Oct 2023

    Manager / Solutions Engineer · Exotel

    Bengaluru. Owned onboarding for enterprise CPaaS deployments and standardised the implementation playbook.

  4. Sep 2020 - May 2021

    Onboarding Manager · WebEngage

    Bengaluru. Marketing-automation onboarding for enterprise customers.

  5. Feb 2020 - Sep 2020

    Enterprise Sales Manager · Exotel Techcom

    Bengaluru. Built the SME segment from one person to a seven-person team. Hiring, comp, targets, forecasting - from scratch.

  6. Oct 2018 - Mar 2020

    Lead Sales Engineer · Exotel

    Bengaluru. 100 dealerships in 30 days. Pipedrive-Exotel API integration via Google Apps Script. Onboarded enterprise customers end to end.

  7. Jun 2017 - Oct 2018

    Inside Sales Engineer · Exotel

    Bengaluru. First job out of college. Exotel Alice Award (Best Performer 2018) in year one. Mechanical Engineering degree, no sales background - closed the gap by getting deep on API integrations.

  8. 2015 - 2017

    Earlier - GoCollab, SkillStreet, BakeryBite, Mascloft

    Mumbai. Marketing, sales, and BD experiments while finishing the MBA-TECH at NMIMS (Mechanical Engineering + MBA in Marketing, 2012-2017). Closed a ₹7-lakh deal at Mascloft cold-calling CEOs at D&B, HP, Taj Group, GroupM.

Get in touch

Want a 30-minute walk-through of any system above?

Pick any agent in the catalogue. I'll show the code, the runbook, and the metric it moves. Same offer for any Ghostbuster mission you'd put in front of me.

Email anuj@scale-with-clarity.com LinkedIn linkedin.com/in/anuj-kapoor-trekker Phone +91 77095 62799
Based Singapore - remote-ready, APAC-anchored

For Alex & Dan - happy to build a 2-week sample mission as part of the interview. Pick a ghost. I'll come back with a diagnosis, a fix, and a number.