Zeeshan Amjad
Software Engineer
across automation, backend, and AI.
I'm a software engineer who builds across test automation, backend systems, and AI-assisted software. I like turning messy manual work into systems that hold up.
About
I'm a software engineer. For the past four years my professional work has been test automation and quality engineering at Arbisoft, and alongside it I build the kind of software I test: multi-tenant SaaS backends, an automation framework, and AI-assisted developer tools.
My day job is test automation, and that work is real. The systems I build on my own run wider and deeper: full-stack and backend-heavy, with their own databases, auth, audit trails, and test suites.
Outside the editor I read a lot, train, and chase questions in philosophy, psychology, and metaphysics. That sensibility shows up here more in how the site feels than in anything I'd say out loud.
Domains charted
Products I've tested at Arbisoft, by category.
- Online learning (MOOC)
Web automation, regression coverage, and workflow validation.
- Online multiplayer gaming
Load testing, regression optimization, and reliability checks.
- Online assessment
Automation coverage for assessment workflows.
- Insurance data extraction
Bot validation, extraction accuracy, and edge-case testing.
The Arsenal
Automation & Quality
Backend & Systems
API & Backend Testing
Performance & Reliability
AI-Assisted Engineering
DevOps & Quality Gates
Data Extraction
The Realm of Systems
The Quality Bastion
The Trade Quarter
The Healer's Ward
The Guild Reach
The Uncharted Frontier
Featured Projects
ForgeLoop
A project-aware control plane that runs the loop from requirement to merge. It drafts tickets and plans, carries out implementation under project rules, holds every change behind deterministic QA checks and a human approval gate, and keeps the evidence: test runs, reviews, audit artifacts, and a memory of earlier decisions.
Sentinel
A tool-agnostic automation framework. You write a flow once in the language of your product and it runs unchanged on web (Playwright) and mobile (Appium) through shared core contracts. Every step emits typed telemetry, and an AI layer reads that signal to tell you whether a failure is a real bug, an infra flake, or selector drift, then proposes locator fixes as diffs you review instead of changing your tests behind your back. Its end-to-end examples also cover scraping and reporting.
Dentaxis
A self-hosted dental practice management system that runs on a clinic's own network by default and is ready for the cloud when needed. Thirty-one modules cover patients, scheduling, charting, clinical notes, treatment plans, surgical logbooks, billing and insurance, payroll, messaging, a patient portal, and an opt-in AI assistant that stays out of clinical decisions.
WhatsApp Order-to-Cash Desk
A multi-tenant SaaS that runs the whole order-to-cash loop. It keeps customers and products, an order board with a strict status flow, payment-proof uploads that staff verify and allocate against open orders, an outstanding ledger with aging buckets, dispatch tracking, follow-up tasks, and reports. WhatsApp Cloud API, screenshot OCR, and bank-statement import are there but switched off by default.
COD Control Desk
A multi-tenant API that scores incoming COD orders and routes the risky ones to a human verification queue, flags likely returns before a brand pays shipping both ways, and reconciles courier payout feeds to surface short payments, missing remittances, and unexplained deductions. It takes three CSV feeds and makes imports previewable and reversible.
Recovery Desk
A deliberately narrow SaaS that turns the weekly expiry-and-stockout spreadsheet into a 30/60/90-day risk dashboard, a costed stockout queue from field visits, and a recovery-action queue with approval gates. At the end of the week it reports the value actually recovered.
Munafasah Desk · BidReady KSA
Two services working together. An Arabic-first admin portal handles tenders and clients, versioned compliance matrices, a document vault, an LLM sector classifier, and the full PDPL surface: consent ledger, data-subject requests, retention with separation of duties, and a residency gate. A separate ingestion pipeline scrapes tender portals, deduplicates, enriches with an LLM, and syncs curated tenders into the portal.
More locations, including work in progress, are out on the world map.
How I Use AI
I use AI as a working partner, not a shortcut. It helps me plan architecture, draft implementations, review diffs, think through test strategy and edge cases, and keep documentation honest. I work with Claude and Claude Code, OpenAI, DeepSeek, and local models, and I keep human judgment, quality gates, and accountability in the loop. What I care about is the workflow: AI that helps plan, build, test, and review software under clear rules.
Architecture planning
Sketching system shape, boundaries, and tradeoffs before any code.
Implementation drafting
Turning a plan into first-pass code that I then review and harden.
Code review
A second pass over diffs for bugs, edge cases, and risk.
Test strategy
Finding the cases worth covering and the ones likely to break.
Product planning
Shaping features and scope for the SaaS work.
Local LLMs
Running smaller models locally to keep agent work affordable.
Cost-aware routing
Sending each task to a model that fits its difficulty and price.
Human-supervised agents
Letting agents do the work behind approval gates and evidence.
“Good software should make things clearer and make repeated work dependable. Mine is shaped by quality engineering, automation, AI-assisted building, and a long habit of taking complex systems apart to see how they hold together.”
The Quest Log
Deepen the agentic SDLC/STLC work
ActivePush ForgeLoop further: better planning, sharper QA gates, and project memory that holds up over time.
Mature the automation framework
ActiveGrow Sentinel's driver coverage and its failure-analysis layer until it earns a team's trust.
Turn builds into products
ActiveTake the multi-tenant SaaS work from working software toward something sellable.
Run more of the stack locally
NextKeep exploring local-first AI so agent workflows do not depend on expensive cloud inference.
Ship scraping and reporting tools
NextHarden the extraction pipelines into reliable, production-grade reporting products.
Grow toward architecture
NextGo deeper on backend and system design, and move toward technical architecture work.
The Chronicle
Software Automation Engineer / SDET @ Arbisoft
Feb 2022 – PresentLahore, Pakistan / Remote
- Build and maintain automated test suites with Playwright, Cypress, TypeScript, and JavaScript across four products: a MOOC platform, an online multiplayer game, an online assessment platform, and a US dental-insurance data-extraction product.
- Automated around 30 cases for the learning platform, two flows for the assessment platform, and five to six load scenarios for the game, covering critical workflows and regression paths.
- Tested 50+ APIs with Postman and automated validation, checking request and response behavior, data, and backend workflow correctness.
- Integrated and maintained automated test runs in GoCD pipelines for CI regression and deployment checks.
- Cut the game's regression run time by about 10% by automating repeated validation flows and steadying execution.
- Investigated and reduced 50+ flaky tests by working through selectors, app state, network calls, logs, timing, and test data.
- Validated the dental-insurance extraction bot for accuracy and edge cases across many insurer websites.
- Improved reliability with reusable fixtures, cleaner test structure, stronger selectors, and retry-aware flows.
Product Support Engineer @ Techlogix
Jun 2021 – Jan 2022Lahore, Pakistan
- Supported enterprise software users by investigating issues, reproducing defects, and coordinating fixes with engineering.
- Analyzed application behavior, user workflows, logs, and data to find root causes and write up clear findings.
- Worked with customers and internal teams to troubleshoot production issues and keep their status visible.
- Built a foundation in defect analysis, product behavior, and customer-facing troubleshooting.
The Record
Resume
The PDF is the ATS-friendly QA / SDET resume. The systems above are the fuller picture.
Certifications
- Responsive Web DesignCompletedfreeCodeCamp
- JavaScript Algorithms & Data StructuresCompletedfreeCodeCamp
- ISTQB CTFL 4.0In progressASTQB
- AWS Certified Cloud PractitionerIn progressAmazon Web Services
Education
- FSc Pre-EngineeringPakistan
- GCE O Levels1 A*, 6 A's, 1 B
Send word
Open to freelance and contract work: automation systems, AI-assisted workflows, backend automation, multi-tenant SaaS tools, and scraping or reporting pipelines.