Azim Haffar/
azimx.dev
Azim Haffar

Software Engineering Student
Backend / Applied AI

Mersin, Türkiye

Internships from 17 Jan 2027

View résumé
GitHubLinkedIn

Azim Haffar · Software Engineering

Backend thinking.
Practical software.

I build with Java and Python—from event-driven backends to applied AI tools. Backend internship at Trendyol. Paid Python/LLM development for Kvote.

Explore my workView résumé
Java / Python

Backend engineering
& applied AI

17 Jan 2027

Internships · 4–12 months
Open to relocation

What survives an outage?Explore the illustrative simulation
OrderFlow
OrderFlow dashboard showing order activity and infrastructure status

Local Kafka recovery experiment

SeatLedger
Deployed SeatLedger portfolio demo showing clearly labeled synthetic accounts and free community events

Deployed synthetic reservation demo

ML Training Inspector
ML Training Inspector showing measured synthetic CPU demo loss, accuracy and gradient charts

Measured synthetic CPU run

01 / Selected work

The interface is only the beginning.

Explore the product, follow a decision, then inspect the evidence.

01 / Featured engineering story

SeatLedger Deployed demo

One seat left.
Two people booking.

A reservation system I built to explore competing requests, expiring holds, and safe retries. Try the product or follow the engineering decisions.

Try the demoEngineering story

Synthetic shared accounts · Free backend may take a minute or more to wake · No personal information

Deployed SeatLedger portfolio demo showing clearly labeled synthetic accounts and free community eventsActual product capture
Actual interface · Synthetic portfolio demo · Real-user release incomplete.
  1. 01 / Protect capacity

    Serialize reservation changes at the event.

  2. 02 / Handle retries

    Persist idempotent outcomes with lifecycle changes.

  3. 03 / Explain the numbers

    Build replayable analytics apart from live inventory.

Verification and demo limitations

18 October 2026 · Deployed portfolio demo · CI: 265 backend, 56 analytics and 28 browser tests passed, zero failures/errors/skips. Synthetic evidence, not production load or real-user validation.

Synthetic shared accounts and events; do not enter personal information. Free backend may sleep, hosted email is disabled, and analytics refresh is manual. Existing hosted events were reported past; booking may be closed. No real-user production launch.

Inspect verification scope

More ways I investigate software.

Recovery, useful feedback, and a view inside training.

Skina local fictional advisor with deterministic demo basket and mocked explanations

02Applied AI systems / Local prototype

Skina

A fictional retail advisor that selects a basket using deterministic rules and explores gated model explanations.

React · TypeScript · Node.js

Evidence & scope
  • Deterministic category and budget selection
  • Gated explanations of selected products
  • Fictional catalog and preserved evaluation misses

Latest retailer prototype is local. Its shopper interface uses mocked explanations; the older public preview is separate. Real-user usefulness and live AI improvement remain unproven.

Case study
Applied ML / Research workbenchMotionTraceEngineering case study ↗

03Applied ML / Research workbench

MotionTrace

A research workbench for tracking an initially annotated target through prerecorded cine-MRI and examining reliability.

Python

Evidence & scope
  • Target tracking and comparator analysis
  • Classical and learned method comparison
  • Negative findings retained

Research and education prototype; not clinical software. Evaluation data are exposed, and publication rights for research assets remain pending.

Case study
OrderFlow dashboard showing order activity and infrastructure status

04Event-driven backend

OrderFlow

Accept work during a Kafka outage, recover pending orders, and inspect stock-safe replay.

Java · Spring Boot · Kafka · PostgreSQL

Evidence & scope
  • Durable work during a Kafka outage
  • Recovery without a second stock deduction
  • 17 tests covering lifecycle and reliability

Local verification, not a load benchmark. Single publisher; no authentication or public hosted demo.

Case studyCodeTry recovery
HireLens offline analysis with illustrative synthetic match and ATS scores

05Full-stack / Applied AI

HireLens

A synthetic offline journey from CV upload and feedback to tracking, history, and deletion.

Python · FastAPI · React · JavaScript

Evidence & scope
  • Upload → feedback → tracking → deletion
  • Desktop + mobile offline journeys
  • Separate scoring and generated advice

Latest improvements are local and unpublished. Mocked providers; live authentication and model quality remain unverified.

Case studyCode
ML Training Inspector showing measured synthetic CPU demo loss, accuracy and gradient charts

06ML tooling / Real-time systems

ML Training Inspector

Live training feedback, checked against a reproducible synthetic PyTorch CPU run.

Python · PyTorch · FastAPI · WebSockets

Evidence & scope
  • Live metrics matched a synthetic CPU run
  • Explicit stop, error and completion states
  • Checkpoint loading verified; resume absent

Synthetic results demonstrate the pipeline, not model quality. Resume is absent; one worker and one shared run.

Case studyCodeExplore run

02 / Experience

Built with teams. Delivered for clients.

Kvote 2 Hjælperen

Freelance Software Developer

Feb 2026 – Apr 2026 · Remote

Delivered a paid Python/LLM study tool with a separate evaluation step and response validation.

Contributions & technologies
  • Delivered a paid Python/LLM study tool that generated practice questions using students' study materials as context.
  • Implemented a separate LLM evaluation step to review relevance, grounding, clarity, and answerability against the source material.
  • Added application-level checks for required fields and response structure, with rejection or regeneration of questions that failed checks.

Python · LLM API Integration

Trendyol

Backend Engineering Intern

Jun 2025 – Sep 2025 · Istanbul, Türkiye · Onsite

Built and demonstrated a Spring Boot coupon-usage service, with PostgreSQL persistence and container-backed testing.

Contributions & technologies
  • Built and demonstrated a Coupon Usage Tracker microservice with Spring Boot, PostgreSQL persistence, and REST endpoints for recording usage, looking up user usage, and checking coupon exhaustion.
  • Wrote unit and integration tests using Testcontainers, documented the service, and packaged it with Docker.
  • Contributed API changes, validation, bug fixes, and tests within existing backend services, and presented engineering decisions to the team.

Java · Spring Boot · PostgreSQL · JUnit · Testcontainers · Docker

03 / Toolkit

A focused toolkit. Real context.

BackendSelect a tool for evidence

APIs and application logic across my internship and personal projects.

Related experience

04 / About

Curious about what happens underneath.

I’m Azim, a Software Engineering student based in Mersin, Türkiye. Backend development is my foundation, with paid applied AI work alongside it.

I enjoy tracing what happens behind an interface—especially when something fails. I build tools that make those hidden processes easier to understand.

Explore that curiosity

Education

B.Sc. Software Engineering

OSTİM Technical University · English-taught
Expected graduation August 2027

English
C1
German
B1 · Goethe speaking module
Turkish / Arabic
Native
My journey so far
  1. Ongoing

    Software engineering studies

  2. Jun 2025 – Sep 2025

    Trendyol internship

  3. Feb 2026 – Apr 2026

    Paid Kvote development

  4. Now

    Reliable systems & useful feedback

A room for learningExplore certificates in 3D
05 / Let’s talk

Good work starts
with a conversation.

Open to paid technical internships from 17 January 2027, with Europe preferred and Türkiye also welcome. Available now for paid remote work that fits university commitments.

azim.haffar@gmail.com
GitHubLinkedInRésumé