HIRA AHMED

Cybersecurity Student & Backend Engineer | Scaling fault-tolerant AI agents & n8n production pipelines

Based in: Karachi, PakistanCurrently: Quality Assurance Lead, TechPotion.aiExperience: 2 years

Hire me

Cybersecurity student at SSUET and Backend Engineer specializing in high-throughput ETL and autonomous AI agents. Proven track record in re-architecting pipelines for 5.6x throughput gains and building stateful AI systems that handle concurrent multi-user workloads. Expert in production-grade automation patterns including durable state management and dead-letter paths.

Experience

Quality Assurance Lead

TechPotion.aiRemote · Malaysia-basedworkSep 2026 – Present

Summary

QA Lead across three simultaneous client contracts, owning end-to-end defect discovery, API-level evidence, authorization, multi-tenant isolation, and business-logic validation.

What I did

  • PharmaConnect · QA Contract
  • Reduced critical defect escape rate by leading QA across 20 end-to-end journeys and 6 user roles, uncovering 11 critical P1 findings out of 80+ total documented defects across 15 modules.
  • Eliminated authorization and isolation regressions by testing authentication, RBAC, tenant isolation, pricing, financial reconciliation, audit logging, and data integrity with full API reproduction evidence.
  • Accelerated remediation by tracing recurring defects to shared root causes, cutting duplicate fix cycles and improving regression coverage across the platform.
  • ToolPotion · QA Contract
  • Improved production listing quality for a 28K+ AI tools directory by documenting 100+ defects across relevance, data integrity, rendering, localization, and leaderboard/category behavior.
  • Protected platform credibility by identifying 190+ dead or fraudulent listings in a targeted production pass and surfacing security risks among live listings.
  • Validated multilingual and publishing workflows end-to-end across two full QA cycles.
  • AI Cloud Academy · QA Contract
  • Eliminated a gamification exploit by identifying and documenting an XP/progress vulnerability, then validating the fix through controlled regression testing across 35 documented defects.
  • Ensured learner-journey reliability by testing progress-tracking, data integrity, and end-to-end business logic across all relevant user flows.

Results

  • Uncovered 11 critical P1 findings out of 80+ total documented defects.
  • Identified 190+ dead or fraudulent listings in a targeted production pass.
  • Eliminated a gamification exploit by identifying an XP/progress vulnerability.
API testingRBACAuthenticationMulti-tenant isolationAudit loggingFinancial reconciliation

Backend & Automation Engineer

TechPotion.aiRemote · Malaysia-basedwork2025 – Present

Summary

Increased pipeline throughput by 5.6× across 28,000+ records by re-architecting ETL into a fault-tolerant parallel system with 50 concurrent workers, dynamic rate limiting, and circuit breakers.

What I did

  • Reduced monthly AWS infrastructure costs by 34% through query optimisation and execution-flow refactoring across Lambda, RDS, EC2, S3, API Gateway, and EventBridge.
  • Achieved ~95% data-extraction success rate by building Python scrapers (BeautifulSoup, Playwright, Selenium) across 10 platforms with 20+ extraction methods and a URL-healing service.
  • Enabled 100% schema-valid LLM output for categorisation, FAQ generation, and 9-language translation by enforcing JSON-mode on a 7-stage status-gated enrichment pipeline via OpenRouter.
  • Shipped 12+ production automation workflows using n8n and Make.com alongside FastAPI, PostgreSQL, Docker, Next.js, and LLM APIs.

Results

  • Reduced monthly AWS infrastructure costs by 34% through query optimization and execution-flow refactoring across Lambda, RDS, and EventBridge.
  • Achieved ~95% data-extraction success rate by building Python scrapers and a URL-healing service that pruned deadlinks and extracted social media metadata from diverse listing structures.
  • Enabled 100% schema-valid LLM output for enrichment by enforcing JSON-mode on a 7-stage status-gated pipeline.
PythonBeautifulSoupPlaywrightSeleniumAWSLambdaRDSEC2S3API GatewayEventBridgen8nMake.comFastAPIPostgreSQLDockerNext.jsOpenRouterJSON-mode

QA Intern

Sadiq.aiKarachi, PakistaninternshipJun 2024 – Aug 2024

Summary

Improved dev-team bug-query efficiency by ~40% by building a RAG-based internal chatbot (Gemini API, Python, Streamlit) to surface defects from tracking data in real time.

What I did

  • Reduced release-blocking defects by documenting 150+ edge-case bugs across web and mobile through manual and automated Flutter testing (unit, widget, golden tests).

Results

  • Documented 150+ edge-case bugs.
Gemini APIPythonStreamlitFlutterUnit testingWidget testingGolden tests

Skills

Technical

PythonPython
n8n
LLM pipelines
AWSAWS
LangGraph
Prompt Engineering
FastAPIFastAPI
Multi-tenant B2B QA
PostgreSQLPostgreSQL
REST APIs
RAG/CRAG
Next.js 15Next.js 15
Dead-letter queues
LangChain
LlamaIndex
GPT-4o
Claude 3.5
OpenRouter
pgvector
Make.com
Async state machines
Rate limiting
Idempotency
SQLSQL
TerraformTerraform
Docker ComposeDocker Compose
Exponential backoff
Groq
FastEmbed
TypeScriptTypeScript
Gemini Flash
Concurrency locks
Z3 SMT solving

Projects

Autonomous Competitor Intelligence & SEO Pipeline

Developerpersonal
Automated competitor research, SEO gap analysis, and brief generation by building a PostgreSQL-backed pipeline with dead-letter queue patterns ensuring 100% data integrity on concurrent updates.
n8nClickUpGeminiPostgreSQL

Enterprise Multi-Tenant QA Matrix

Authoropen_source
Open source framework for testing enterprise multi-tenant applications.
QAMulti-tenancy
View the code

Autonomous Cloud Compliance & AI Auditing Engine

Lead Developerpersonal
Cut questionnaire turnaround by ~90% by engineering an autonomous engine that auto-answers 250+ question enterprise reviews (CAIQ, SOC 2, SIG) from live AWS infrastructure evidence. Eliminated LLM hallucinations (100% citation enforcement) via a 6-node LangGraph CRAG state machine with Llama-3.1-8B relevance scoring and an automated Critic Agent verifying real AWS ARNs. Achieved 0.95 semantic relevance score at $0 embedding cost using PyMuPDF + FastEmbed all-MiniLM-L6-v2 stored in Supabase vector(384); processed 2,000+ question workloads with zero timeout failures.
LangGraphCRAGLlamaIndexSupabase pgvectorFastAPIn8nAWSLlama-3.1-8BPyMuPDFFastEmbedall-MiniLM-L6-v2

n8n Production Resilience Patterns

Authorpersonal
Open-source repository showcasing production-grade n8n patterns including durable state management, retries with exponential backoff, and dead-letter/replay paths for fault-tolerant automation.
n8n
View the code

Autonomous AI Email Agent

Developerpersonal
Eliminated manual screening overhead by building a stateful email agent with thread-level memory for automated candidate screening and technical Q&A, supporting concurrent queries from multiple users with explicit branching for malformed inputs.
n8nGmail APIGroqRAGGoogle Sheets

Education

B.S. Cybersecurity

Sir Syed University of Engineering and Technology (SSUET)Cybersecurity2024 – Dec 2028

Societies & activities

Top 10 Finalist (Blue Team), Ignite National Hackathon

Recognition

Awards

Top 10 Finalist (Blue Team)

Ignite National Hackathon

Competed among 200+ teams

Certifications

ISC2 Certified in Cybersecurity (CC)

ISC2Credential 1dd96aa4-0610-46f1-8751-1d17853479aa2025

Patents & publications

Sentinel-Mesh: Formally Verified Remediation of Cloud Misconfigurations

research paperTCC-2026-08-06932026

Neuro-symbolic framework combining LLM-generated Terraform remediation with Z3 SMT formal verification. Evaluated on CloudFix-Bench (105 AWS Terraform cases): achieved 83.81% remediation rate (88/105), 0.0% security regression, and 29 formal proof certificates. Under review at IEEE Transactions on Cloud Computing.

Read it

Social impact & activities

Verified Peer Reviewer

IEEE Accessother

Completed 5 reviews, indexed in Web of Science / ORCID

What drives the work

Motivations

  • AI Engineering
  • Agentic Workflows
  • B2B SaaS QA
  • Data Engineering
  • Early-Stage Startups
  • Production LLMs
  • Workflow Automation
Hire HIRAReach out about a role, a contract or a conversation.For recruiters

HIRA's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis