Open to remote roles · Multan, Pakistan · UTC+5

I find out why abusiness systemis broken —then ship the fix.

CRMs, web apps, databases, integrations and AI agents — built with Claude Code as my co-engineer, and run with the client in the room.

Fasih Sajid · Project Manager & Guest Experience Supervisor — Creative Projects
GoHighLevel Certified Admin · BSc Computer Science (in progress)

See the work → Download CV Book a call
fasih@diagnose — zsh

A diagnosis run: 98 of 99 deals went to one rep; the causes were a round-robin setting and an assign step; five workflows shipped.

6shipped systems, written up below
$576Krecurring revenue from a department I founded
4.9CSAT across 7,000+ client calls
5,000+GoHighLevel clients supported

Looking for: forward deployed engineer · AI solutions / implementation engineer · AI automation engineer · applied AI · technical PM

01 — About

The person behind the work

I'm a 21-year-old builder from Multan, Pakistan, and today I'm Project Manager & Guest Experience Supervisor at Creative Projects (HL Pro Tools) — the person clients reach when something has gone wrong and needs to be put right.

In under three years I went through eight roles, from chat specialist to supervisor, while founding a department that didn't exist, supervising a $150K+ launch and leading support floors.

Along the way I started building: Chrome extensions, Supabase apps, account monitors and a whole Claude Code setup of skills, hooks and integrations. I'm studying Computer Science at NCBA&E and use everything I learn on real systems the same week.

Let's work together LinkedIn ↗
Fasih Sajid
Location
Multan, Pakistan · UTC+5
Education
BSc CS — NCBA&E · 2023–present
Languages
English C1 · Urdu (native)
Status
● Open · full-time / contract

02 — Selected work

Six systems, told the way I'd tell them on a call

What was broken, what the data said, what actually caused it, and what shipped. Client names are withheld; the numbers are real.

Diagnosis · event rental · live paid traffic

98/99deals owned by one rep
2separate root causes
5workflows shipped

Why a client's paid leads never moved

Situation

A GoHighLevel account running live Google and Meta spend, two sales reps — and one of them was getting everything. The owner was about to rebuild the funnel to fix it.

What the data said

98 of 99 opportunities belonged to one rep, and 94 of 99 sat in the first pipeline stage despite confirmed appointments. Paid traffic landed, then went nowhere.

Actual cause

Two faults everyone had merged into one. The round-robin calendar optimised for availability, not equal distribution. And a workflow assign step reassigned contacts that already had an owner, overwriting the second rep on every run. The funnel was fine.

What shipped

Two calendar fields fixed and verified by re-reading the object key by key. The assign step corrected. Five workflows authored and imported so deals now move themselves as appointments are booked, held and quoted.

  • GoHighLevel
  • REST API
  • Round-robin routing
  • Pipeline automation

Reverse engineering · own product

404what the public API returns for a workflow
v5.9shipped versions of the extension
10test suites passing

Making GoHighLevel workflows portable when the API won't

Problem

GoHighLevel has no supported way to read or write a workflow. GET /workflows/{id} returns 404 even to a fully authorised token, so automations can't be versioned, reviewed or moved between accounts.

What I worked out

How the builder actually stores workflows: two config styles that must never be mixed in one account, two condition grammars, and a silent-drop trap where an invalid step vanishes on import with no error.

What shipped

Axiom MVP for HighLevel, a Chrome extension that exports workflows, forms, surveys and email templates to JSON and imports them back — always as drafts, never overwriting. Cross-account moves remap pipelines, stages and fields by name.

Judgement call

Before selling it, I read HighLevel's terms. The read found the internal-API approach wasn't safe to sell as built, so it's paused pending a lawyer's review instead of shipped to customers.

  • Chrome MV3
  • JavaScript
  • JSON schema
  • GitHub Actions

Product page ↗

Full-stack app · own product

$582Kopen deals found at risk in one account
9health checks
37automated tests

A watchdog for agency client accounts

Situation

I read 186 public reviews of GoHighLevel to find a real gap. The top pain: an agency's client account breaks or goes quiet, and nobody notices until the client does.

What it checks

Leads stopping, unanswered leads and missed calls, speed to lead, forms going quiet, changed or erroring workflows, calendars, phone numbers — and revenue at risk: open deals waiting on a reply or going cold.

Design decisions

Read-only, always. A failed read shows as "unknown", never a clean zero. Deals over 20× the median are left out and flagged, which caught a $1M placeholder in a pipeline where the typical deal is $4K.

What shipped

A dashboard with a recovery list that deep-links into each contact, acknowledge and snooze, and white-labelled weekly client reports. Every stage count and dollar total matched the GoHighLevel screens. Runs for five agency accounts.

  • Node.js
  • Supabase
  • Vercel
  • GoHighLevel API
  • Slack

AI tooling · my own setup

282sessions logged automatically
19custom Claude Code skills
~22Ktokens saved per session

A Claude Code setup that remembers, files and guards itself

Problem

AI sessions forget everything. Decisions, client context and hard-won gotchas were being re-learned every time — and secrets pasted into a session could end up in notes.

What I built

Session hooks that summarise every Claude Code session into an Obsidian vault and file what it produced: people, projects, decisions, insights and tasks. A catch-up sweep re-logs anything a hook missed.

A bug worth telling

Saving "the current session" picked the newest transcript by file time. With parallel sessions that's arbitrary, and it archived the wrong ones. Fixed by resolving the session by ID, and refusing to guess when it's ambiguous.

Guardrails

Credential-shaped strings are redacted before anything is written, and a scan of every note against the credential store found and fixed two leaks the patterns missed. Plugin overhead was audited and cut.

  • Claude Code
  • Hooks
  • Skills
  • MCP
  • Bash · Python
  • Obsidian · Git

Client-facing · Creative Projects

39people in a one-call team snapshot
~6sto build it
3ways every logged hour is split

Turning client escalations into facts, credits and replies

Situation

Clients of a virtual-assistant service escalate with "the hours are wrong" or "nothing's been done". The answer is spread across a client thread, handoff notes, a work queue and time logs — in a tool whose own comment counts can't be trusted.

What I built

Two read-only Claude Code skills. One audits a task, client or whole agency: reply times, promises kept, estimate history, missing end-of-day notes, and where internal notes contradict what the client was told.

The ledger

The second turns an audit into an hours ledger: original scope, client-driven change, and team rework — and flags overrun the client never authorised. It drafts the credit recommendation and the client reply.

Outcome

One audit found a "refining the landing page" update for a site that had never been accessed — the credentials pointed at another client's site. Another led to a full refund plus honoured hours. Facts first, then the apology.

  • Claude Code skills
  • JavaScript
  • Read-only API calls
  • Escalation handling

Web app & database · agency

6Postgres tables behind row-level security
0frameworks or build steps
3views: table, subtasks, kanban

Replacing an agency's task spreadsheet with a real app

Situation

A small marketing agency ran client work from a shared spreadsheet: no owners on subtasks, no overdue view, no way to tell what was blocked.

Security model

Access is enforced in the database, not the page. Every row-level policy checks a members list. Only a server-side function holding the service key can change who's on it.

What shipped

Tasks with inline status and priority, subtasks with owners and due dates, a cross-task subtask view, a kanban board, search, filters, overdue counters and custom columns without migrations.

Why no framework

Plain ES modules on Supabase and Vercel. Nothing to build, nothing to upgrade, and the next person can read all of it in an afternoon.

  • Supabase
  • Postgres RLS
  • Edge functions
  • Vanilla JS
  • Vercel

Also shipped

  • Approval-gated outreach dashboardNext.js + Supabase on Vercel. Leads from three sources in one queue; nothing sends without a human approving it.
  • Social publishing pipelinePython + Supabase + the Zernio API. 150 posts staged, 91 LinkedIn posts published, and a reconcile script that catches silent failures.
  • WordPress → GoHighLevel site migrationAudited a live agency site and rebuilt 15 pages as import files, each checked against the original.
  • Site performance passaxiom-arc.com: largest paint 3.4s → 1.6s, layout shift 0.207 → 0.01, structured data and a real 404.
  • Team utilisation audit48 people, 911 hours tracked against 1,256 clocked. Found the export itself was unreliable, and a 7.8% gap between two time sources.
  • AI receptionist & crew schedulingA cleaning company's GoHighLevel build: conversation AI, four workflows, calendar and phone routing.

03 — How I work with AI

Claude does the typing. I own the judgement.

Most of what's above was built with Claude Code. These are the rules it works under — the same ones I'd bring to a customer's systems.

  1. Read before write.Every change starts by reading the current state through the API — never from memory or a note that might be stale.
  2. Plan, then wait for a yes.Anything that touches a live account gets a written plan first: what exists, what changes, what stays untouched, how it's checked.
  3. Never edit what's running.New behaviour goes in duplicated assets. Imports land as drafts. Nothing is published by the tool.
  4. Verify by ID, not by name.Accounts and records are confirmed by their ID before anything is written. Names collide; IDs don't.
  5. A failed read is "unknown", never zero.Every count is re-derived before it's reported. An empty list is a question, not an answer.
  6. Humans approve anything that sends.Outreach, client replies and messages are drafted by AI and sent by a person.
  7. Secrets stay in one place.Credentials live in a single local store; logs are redacted and scanned for anything that slipped through.

04 — Capabilities

What I build and run — with the proof

Every group links to the case study that backs it.

Web apps & software

Next.js, TypeScript, plain JavaScript, Node, Chrome extensions.

030602

Databases

Supabase / Postgres, row-level security, edge functions, schema design.

0603

APIs & integrations

REST APIs, webhooks, GoHighLevel, Slack, Zernio, Google Workspace.

0301

Automation

GoHighLevel workflows, n8n, Zapier, Python and shell scripts.

0104

AI agents, skills & MCP

Claude Code skills, hooks, subagents, MCP servers and integrations.

0405AI

Reverse engineering

Working out undocumented formats and internal APIs, then making them safe to use.

02

Data analysis

Finding the number that explains the problem — and checking the data source itself.

010305

Clients & teams

Discovery calls, escalations, SOPs, coaching, and leading support floors.

05XP
  • Claude Code
  • Claude API
  • MCP
  • Next.js
  • TypeScript
  • Node.js
  • Supabase
  • Postgres
  • Vercel
  • Netlify
  • GitHub Actions
  • GoHighLevel
  • n8n
  • Zapier
  • Make
  • Python
  • Slack
  • Airtable
  • Stripe
  • Twilio / A2P

05 — Experience

Eight roles in under three years

All remote, all at HL Pro Tools (Baam) — from chat specialist to running client escalations.

  1. Aug 2026 — Present

    Project Manager & Guest Experience Supervisor Current

    Creative Projects · HL Pro Tools (Baam)

    • Own client escalations for the virtual-assistant teams: audit the full task history, find what actually went wrong, and recommend the credit, refund or recovery plan.
    • Built read-only Claude Code audit tooling that checks reply times, promises, estimates and handoff notes, and snapshots a 39-person team's workload in one call.
    • Turn audits into hours ledgers and client replies that stand behind every logged hour.
  2. Dec 2025 — Aug 2026

    Guest Support Supervisor

    HL Pro Tools (Baam)

    • Ran the support floor across chat, video, tickets and email; held CSAT at 4.7+ by redesigning coaching cadence and audit cycles.
    • Audited team utilisation reporting and caught that the source export itself was unreliable.
    • Wrote the handover playbook for the incoming supervisor.
  3. 2025

    Launch Supervisor & Project Manager — Search Atlas

    HL Pro Tools

    • Supervised 400+ onboarding calls in 1.5 months for a launch that brought in $150K+.
    • Managed a conversion team to a 75% close rate with call-flow playbooks and objection handling.
  4. 2025

    Enterprise Supervisor

    HL Pro Tools

    • Routed enterprise escalations and high-value accounts across support and account teams.
  5. 2023 — 2025

    Founder & Lead — Mastermind Tech Admin Department

    HL Pro Tools

    • Founded a department that didn't exist: $24,000/month in new recurring revenue, sustained for two years (~$576K).
    • 4.9 CSAT across 7,000+ client calls; wrote the operating model, SOPs and escalation paths from a blank page.
  6. Jul 2024 — Jul 2025

    Team Lead · Senior Technical Expert · Senior Chat Specialist

    HL Pro Tools · promoted 3× in 12 months

    • Certified GoHighLevel admin support for 5,000+ clients: funnels, automations, A2P, email, Shopify, migrations and agency setups.
    • Built internal Claude and Zapier tools to speed up GoHighLevel troubleshooting.
  7. Before 2024

    Remote Truck Dispatcher

    US logistics firms · remote

    • Real-time freight dispatch for US carriers — where the remote discipline and client communication started.
Download the full CV Book a call

06 — Recognition

Earned recognition

2024

GoHighLevel Certified Admin

HighLevel · issued Nov 2024

The platform's full admin certification, earned while supporting 5,000+ client accounts.

2023–25

Founding Lead — Mastermind Tech Admin

HL Pro Tools · internal recognition

A brand-new department: $24K/month in recurring revenue (~$576K over two years) at a 4.9 CSAT across 7,000+ calls.

2024

Promoted 3× in 12 months

HL Pro Tools · career track

Chat Specialist → Senior Technical Expert → Team Lead, on CSAT, throughput and client outcomes.

2025

$150K+ Search Atlas launch

HL Pro Tools · launch performance

Supervised 400+ onboarding calls in under 1.5 months while holding quality.

07 — Testimonials

Real words from real clients

From colleagues and the clients I've supported — unedited.

★★★★★
“Fasih consistently brought a strong work ethic, innovative thinking and a results-driven approach to every project. His technical expertise and problem-solving skills played a crucial role in streamlining our processes and driving team success. A quick learner, adaptable, and always willing to go the extra mile.”
Taha KhanAI Product Builder & Ops Leader · HL Pro Tools
★★★★★
“Fasih is amazing at GoHighLevel, consistently demonstrating his ability to resolve issues instantly. His strong technical background lets him tackle complex problems with ease. His quick troubleshooting and in-depth knowledge make him an invaluable resource for our team and our clients.”
Mahrukh IqbalDirector of Ops & Client Success · GHL & AI
4.9Live post-call CSAT across 7,000+ client calls
★★★★★

“This interaction couldn't have been better. Fasih was very efficient and knew exactly what my issue was — step by step how to resolve it. Best customer service I've ever received. 10/10.”

James R.Verified client
★★★★★

“Fasih is absolutely amazing — so patient, polite, professional. Spent a long time with me until the job was done! Customer service like no other I have experienced. Thank you Fasih, you're a star!”

Sandra H.Verified client
★★★★★

“Fasih had such calm and welcoming energy. Amazing energy, positive attitude, I didn't feel rushed. I felt truly supported and encouraged through the process. I appreciate you!”

Karen W.Verified client
★★★★★

“Incredible — Fasih provided so much help, patience, clarification and most importantly encouragement! Attentive to my needs. It was above and beyond for me. Thank you!”

Rachel P.Verified client
★★★★★

“Fasih not only resolved the issue but took the solution to the next level — made me a video explaining it and why it was duplicated. Fasih gets the 5 stars for going beyond just resolving.”

Jennifer T.Verified client
★★★★★

“Fasih was so patient with a frustrating issue and resolved it completely after quite a bit of time. I couldn't have done it without him!”

Steven K.Verified client
★★★★★

“Fasih helped me tremendously today! He understood my questions and guided me to the proper section inside the Mastermind Business Hub. He understands the platform extremely well.”

Eric D.Verified client

08 — Credentials

Certified, and still studying

HighLevel Certified Admin badge issued to Fasih Sajid on November 6, 2024, credential 121710512
  • GoHighLevel Certified AdminHighLevel · issued Nov 2024 · verify on HighLevel ↗ · Credential.net ↗
  • BSc Computer ScienceNCBA&E, Multan · 2023 — present
  • A-LevelsBritain International College Network · 2022
  • O-LevelsBritain International School System · 2020

09 — Let's talk

Have a system that isn't doing what it should?

I'm open to forward deployed, AI solutions and automation roles — full-time or contract, fully remote. A 20-minute call is the quickest way to see if it fits.

📅 Book a call Download CV