Player 1 · Hamza Mehmood Open to .NET & backend roles · Vienna, VA or remote

Software that keeps working when things go wrong.

I build .NET backends: event-driven services, fast SQL, and APIs that are tested until they're boring. Four years across enterprise fintech, healthcare billing and Azure integration platforms.

Enter the arena ↓ Email me GitHub LinkedIn
Achievement unlocked180s → 32sA reporting module, after profiling execution plans
Achievement unlocked91%Fewer production defects on a three-client platform
Achievement unlocked30–40Production incidents a month I owned across four systems
Achievement unlocked640+Automated tests across 18 public projects, all in CI
Stage 1 · The arena

Try to break it.

A live model of the retry and circuit-breaker pattern from my Resilience Toolkit. Make the dependency unreliable, send traffic, and watch the breaker stop the pile-up, then heal itself.

Breaker closed · healthy
0Served
0Retried
0Shed instantly (no wasted call)
0Gave up
  • Press “Take it down”, then “Send 10 requests”.

Simulation only, but the logic mirrors the tested implementation in resilience-toolkit-dotnet (retry with jitter, circuit breaker, timeout, bulkhead).

Stage 2 · Quests

The parts that can't fail.

Integrations, reporting and payments-style workflows: the systems where a slow query or a lost message costs real money. Here's how I approach them.

Ability · Production integrations

Four systems, one platform.

Azure · Service Bus · retries · dead-letter
Ability · Performance

180 seconds down to 32.

Execution plans · indexed views · stored procedures
Ability · Correctness

Proven under contention.

Idempotency · outbox · optimistic concurrency
Quest complete

Consultant .NET, Systems Limited

PartnerLinQ platform · May 2024 – Jul 2025

Primary production responder for a platform integrating four downstream systems.

Inside the work
Problem
Partner systems failed in different ways, and recurring incidents ate the team's week.
What I did
Built event-driven services on Azure Functions, Logic Apps, API Management and Service Bus with retry, idempotency and dead-letter handling. Added logging and telemetry, and traced each incident to root cause instead of patching symptoms. Managed environments with Terraform and shipped through Azure DevOps.
Result
30–40 incidents a month handled at 70% SLA compliance, and 30% fewer recurring failures.
Quest complete

Software Engineer, NETSOL

Lease & finance platform · Jun 2022 – Apr 2024

Owned 10+ modules end to end for an enterprise platform serving three clients.

Inside the work
Problem
A core reporting module took 180 seconds, so operations teams worked around it.
What I did
Profiled the execution plans, removed table scans across two large joins, and redesigned the stored procedures with indexed views. Led code reviews and mentored junior developers.
Result
180 s to 32 s, production defects down 91%, and sprint delivery steady at 94%.
Side quest · ongoing

Reference implementations

Public, tested, verified in CI

The patterns I use at work, rebuilt in the open with tests that try to break them.

Inside the work
Idempotency
Eight concurrent duplicate requests create exactly one order.
Concurrency
Ten parallel withdrawals against a balance that covers five: exactly five succeed, and the balance never goes negative.
Delivery
A transactional outbox with competing dispatchers and an idempotent inbox, so events survive a crash and never run twice.
Architecture
One checkout system built as a modular monolith and as four .NET services. The same scenario suite passes against both, and a test fails the build if a module reaches into another module's code.
Live demos
An animated cafe that shows why systems write orders down first, and an A/B test analyzer that shows why peeking at results gives false wins.
Stage 3 · Boss fights

Read the source.

Every project has a README, automated tests and CI. None contains employer code. The health bar is the test suite, and it is full.

Hold My Seat: seat map with two buyers
Boss · The Double BookingHold My SeatTwo buyers, one seat. Holds expire, purchases retry safely, and nobody is ever double-booked.
HP42 tests
Fight it live →
 Pipeline CRM: deal board with stages
Boss · The Duplicate CustomerPipeline CRMLeads that never duplicate, a strict deal pipeline, an honest forecast, and a sync to another system that survives outages.
HP142 tests
Fight it live →
 Till point of sale register screen
Boss · The Double CountTill: Offline POSSell offline, sync later, and every sale lands exactly once. Refunds and drawer balancing too.
HP61 tests
Fight it live →
 Coffee Shop Rush: animated cafe scene
Boss · The Lost OrderCoffee Shop RushCrash the barista and see why real systems write orders down first. No jargon.
HP23 tests
Fight it live →
 A/B Test Analyzer calculator
Boss · The False WinnerA/B Test AnalyzerDid the variant really win? Significance, sample size and why peeking misleads.
HP24 tests
Fight it live →
 SaaS product analytics report
Boss · The Silent Drop-offSaaS Analytics in SQLFunnel, retention and channel quality in plain SQL, with the query behind every number.
HP10 tests
Fight it live →
Hold My SeatSeat booking in C#: expiring holds, all-or-nothing reservations and idempotent purchases, proven with real multi-threaded contention tests.42 tests↗ Till: Point of SaleOffline-first register in C#: split payments, discounts, refunds, shift reports, and exactly-once sync after an outage.61 tests↗ Pipeline CRMCRM-style project: duplicate-proof leads, strict deal pipeline, forecast, OData-style filters, and a versioned sync to another system.142 tests · live↗ Multi-tenant SaaS APITenant isolation at the query and write layers, roles, and per-tenant rate limits: one customer can never see another's data.28 tests · live↗ Webhook GatewaySigned webhooks with replay protection, retries, dead letters, and a receiver that processes each event once.68 tests · live↗ Job SchedulerCron scheduler with retries, missed-run and overlap policies and dead letters, on a clock you control.76 tests · live↗ Monolith vs MicroservicesOne checkout system deployed as a modular monolith and as four .NET services; one scenario suite proves both behave the same.34 tests↗ OrderFlowIdempotent order API with retries, backoff and dead-letter handling.17 tests · live↗ Transactional OutboxReliable event publishing with competing dispatchers and an idempotent inbox.19 tests · live↗ Event-Sourced LedgerCQRS and optimistic concurrency: parallel withdrawals can never overdraw.26 tests · live↗ Claims APIClean Architecture, JWT and a role-based claim workflow enforced in the domain.22 tests · live↗ Resilience ToolkitRetry with jitter, circuit breaker, timeout and bulkhead. Powers the arena above.23 tests · live↗ API Rate LimiterToken bucket and sliding window as ASP.NET Core middleware; 1,000 concurrent callers never exceed the limit.25 tests · live↗ XML Mapping ServiceXSD validation and sandboxed XSLT with XXE protection.11 tests · live↗ SQL Performance LabQuery tuning proven by identical results and query plans.6 tests↗

Side quests · full-stack Angular + .NET

Claims Portal screenshot
Claims PortalAngular front end for the Claims API: JWT login, guards, role-based workflow.
 Employee Directory screenshot
Employee DirectoryAngular + minimal APIs with live async validation and server-side paging.
 Inventory Manager screenshot
Inventory ManagerFull-stack CRUD with search, sorting, paging and stock adjustments.
Stage 4 · Bestiary

Bugs defeated.

Real bugs that surfaced while building these projects, caught by tests, and what each one taught me.

SQLite · EF Core The Timezone Ghost

The timestamp with no timezone.

SQLite drops DateTimeKind, so dates came back without a Z and browsers read UTC as local time. A value converter fixed it, and a test on the raw JSON keeps it fixed.

ASP.NET Core · records The Silent Validator

Validation that silently didn't run.

[Required] on a positional record lands on the constructor parameter, but minimal APIs validate properties. Blank names sailed through until [property: Required].

Concurrency The Overdraft Hydra

Ten withdrawals, balance for five.

Optimistic concurrency with reload-and-retry means the rules always run against the latest state. Exactly five win, and the balance never goes negative.

SQL The Phantom Speedup

Measure, then change.

Tuning is only real if results stay identical. Each optimization in the SQL lab is asserted against the original query's output and plan.

Stage 5 · Campaign

Learning by shipping.

From enterprise .NET to research on how AI systems get evaluated.

  1. Aug 2026 – now
    Volunteer Research Assistant · Texas Tech UniversityBuilding LLM-as-a-judge methods that score generated output against rubrics.
  2. Feb – Aug 2026
    Graduate Assistant · TTU OnlineShipped a production chatbot on a vector database (under 3 s, 30% lower API cost) and messaging and monitoring agents.
  3. May 2024 – Jul 2025
    Consultant .NET · Systems Limited (PartnerLinQ)Azure integration platform, production support, Terraform and Azure DevOps.
  4. Jun 2022 – Apr 2024
    Software Engineer · NETSOL TechnologiesEnterprise lease and finance platform for three clients.
  5. Oct 2021 – Apr 2022
    .NET Core Developer · Hawk Logix80+ REST endpoints for a US medical billing platform.
  6. 2017 – 2026
    M.S. Computer Science, Texas Tech · B.S. Software Engineering, National Textile UniversityAdvanced databases, algorithms, software verification and project management.
Stage 6 · Loadout

What's in the inventory.

Grouped by how I use them: what I reach for every day, what I use well, and what I'm still levelling up.

Main weapons

Daily drivers
C#.NET 8ASP.NET CoreWeb APIEF CoreSQL ServerxUnitAzure FunctionsService Bus

Support gear

Strong working knowledge
DockerTerraformAzure DevOpsGitHub ActionsPostgreSQLAngularBlazorPythonOAuth 2.0 / JWT

Leveling up

Newer territory
LLM APIsRAGVector databasesLLM-as-a-judge evaluationProduct analytics
Stage 7 · Next level

Ready to play?

A role, a question, a hard problem? Let's talk.

hamza.mehmood46@gmail.com LinkedIn GitHub Résumé ↓