Hira Ahmed

Production proof

Figures from production automation and QA work. Each one is explained in the case studies.

• 28,000+Production recordsProcessed by a production AI enrichment pipeline
• 5.6xPipeline speedupFault-tolerant parallel processing across 28K+ URLs
• 34%AWS cost reductionMonthly infrastructure cost, via query and execution-flow optimization
• 678+QA defects documentedIdentified and documented across QA engagements and production testing
• 20End-to-end QA journeysOn PharmaConnect, a multi-tenant B2B platform
• 105CloudFix-Bench patternsAWS Terraform misconfiguration benchmark for Sentinel-Mesh

Professional experience

1. Sep 2026 – Present · Remote

   Quality Assurance Lead · TechPotion.ai

   Leading QA across current production B2B software engagements.

   • Lead QA on PharmaConnect: 20 end-to-end journeys across 6 user roles, 80+ defects with API-level reproduction evidence, and 11 critical P1 findings identified during enterprise QA and security testing.
   • Tested authentication, authorization, multi-tenant isolation, pricing, financial reconciliation, and data integrity.

   PharmaConnect

   Multi-tenant B2B pharmaceutical distribution and logistics platform.

   • Lead QA across 20 end-to-end journeys and 6 user roles.
   • 80+ defects documented with API-level evidence across 15 modules; 11 critical P1 findings.

   ToolPotion

   Production AI tools directory with 28K+ listings.

   • Logged 100+ defects across two QA cycles.
   • Identified 190+ dead or fraudulent listings in one pass using targeted search queries.
   • Identified security risks among live listings, including phishing, malware, and gambling-redirect listings.
   • Traced recurring defect clusters to shared root causes to reduce duplicate remediation work.

   AI Academy Cloud

   AI education and certification platform.

   • Documented 35 defects during QA and content-system testing.
   • Identified an XP/progress exploit and multiple progress-tracking issues.
   • Verified core platform functionality under load.

2. Jan 2025 – Jan 2026 · Remote · Malaysia-based company

   Backend & Automation Engineer · TechPotion.ai

   Built and ran the production AI pipeline, automation, and cloud systems behind a 28,000+ listing AI tools directory.

   • Built Python scrapers with BeautifulSoup, Playwright, and Selenium across 10 platforms; 20+ extraction methods with a reported 95% success rate across 28K+ URLs.
   • Reworked ETL into a fault-tolerant parallel system with 50 concurrent workers, dynamic rate limiting, and circuit breakers, reporting a 5.6x speedup.
   • Built a production AI / LLM enrichment pipeline: status-gated stages for categorization, descriptions, FAQ generation, and 9-language translation through OpenRouter (Gemini Flash Lite), with structured outputs and failed stages recorded for retry instead of advancing.
   • Reduced monthly AWS infrastructure costs by 34% through query and execution-flow optimization.
   • Built 12+ Make.com AI automation flows and shipped a multilingual Next.js SSR frontend with FastAPI services.

3. Jun 2024 – Aug 2024

   QA Intern · Sadiq.ai

   Manual and automated QA across web and mobile applications.

   • Identified and documented 150+ bugs, working with developers on prioritization.
   • Worked with Flutter unit, widget, and golden testing; built automated test scripts for core features.
   • Developed a Python, Streamlit, and Gemini-based chatbot.

Featured case studies

How the numbers above were produced: the problem, the architecture, and the result.

What I build and test

AI systems and automations for production, and the QA that keeps them trustworthy: LLM pipelines, agents, n8n workflows, and testing for B2B SaaS.

• AI Automation

  Workflows that connect your tools, APIs, databases, and LLMs, and keep working when inputs are messy or services fail.

  Python · n8n · Make.com · PostgreSQL · AWS

• n8n Automation

  Self-hosted or cloud n8n workflows with explicit state, error routing, and dead-letter queues, so failures are caught instead of lost.

  n8n · PostgreSQL · Docker · ClickUp · Slack

  Production Resilience Patterns → (opens in a new tab)

• AI Agent Development

  Agents that keep context across conversations, answer from verified sources, and hand off to people when they should.

  LangGraph · CRAG · LlamaIndex · pgvector · FastEmbed

• API Integrations

  Reliable connections between SaaS tools, databases, and AI providers, with authentication, retries, and logging built in.

  REST APIs · Webhooks · OAuth · n8n · Make.com

• B2B SaaS QA

  Pre-launch and production QA that goes past the UI: journeys, network/API evidence, controlled reproduction, and root cause.

  Browser DevTools · Network/API inspection · curl · Flutter testing · Defect matrices

• API Testing

  Testing the API behind the interface: auth, permissions, payloads, status codes, and what happens to state when things fail.

  HTTP · REST · curl · Browser network inspection · JSON schema checks

• Multi-Tenant Testing

  Checking that each tenant, role, and session only sees and changes what it should.

  RBAC matrices · API testing · Session/token checks · Data-integrity checks

  QA Matrix → (opens in a new tab)

• Business Logic Testing

  Finding the defects that pass every form validation but still cost money: pricing, invoicing, reconciliation, and state rules.

  Exploratory testing · Controlled reproduction · Fault injection · API testing

Selected projects

Automation systems with public demos or code. See all projects.

Research

Cloud-security research on combining LLMs with formal verification, plus peer-review work for IEEE Access.

Technical capabilities

Tools I use in production across automation, backend, cloud, and QA.

How I work

I build AI systems and I test them. Doing both means reliability is designed in, not bolted on.

1. Build

   Design the AI system or automation around real data and real failure modes: where state lives, what gets retried, and where failed payloads go.

2. Integrate

   Connect models, APIs, databases, and business tools, with structured outputs so downstream steps can trust what they receive.

3. Test

   Test it the way I test client platforms: end-to-end journeys, edge cases, malformed inputs, and forced upstream failures.

4. Verify

   Check the stored data and the API behavior, not only the screen, and verify fixes with regression checks.

5. Deploy

   Ship to production with documentation and a Loom walkthrough, so your team knows how it works and what to do when something breaks.