Orçun Candan — Senior Software Engineer
info@orcuncandan.com · Dubai, UAE
Experience
AIMultiple — Full Stack Developer
Production platform engineering, applied AI and automated data workflows.
- Build PHP and JavaScript/Next.js systems serving 500,000+ monthly users, across performance, SEO, usability and reliability.
- Introduce AI workflows for multilingual content, model and GPU benchmarks, review analysis and development automation.
- Build scraping and validation pipelines collecting structured data from hundreds of web sources.
- Work with AWS and caching; previous performance work reduced render times by about 35%.
Turquoise Choice — Engineering Lead / Full Stack Developer
Architecture, technical leadership and gradual legacy modernization.
- Led architecture, code reviews and critical application development while supporting developers and QA.
- Designed a compatibility layer between legacy WordPress and Laravel services for migration without interrupting the existing product.
- Improved delivery with Azure Web Apps, Azure DevOps and CI/CD; reduced regressions through stronger reviews.
FTRZ Agency — Full Stack Developer
WordPress development, API solutions and AI-assisted content workflows.
- Built WordPress plugins, themes and API solutions, including an early ChatGPT-assisted content workflow.
- Optimized mobile performance and worked on an internal CodeIgniter-based CRM plugin.
Ozkula Internet Hizmetleri — Full Stack Developer & Systems Engineer
Web applications, accounting integrations and production infrastructure.
- Built REST integrations automating more than 10,000 e-invoice transactions monthly.
- Developed Laravel applications, support dashboards and monitoring across 100+ servers.
- Handled Linux/MySQL recovery and web-service troubleshooting; automated recovery work with Bash.
Atlantis Dekoratif ve Sanatsal Ürünler — Full Stack Developer
E-commerce, stock and order management for a large catalog.
- Built systems for 7,000+ products with 15+ variations per product.
- Implemented dynamic product images and later moved parts of rendering to browser-side JavaScript.
LG Electronics Türkiye — Application Developer
Interactive Facebook contest application development.
- Developed a PHP/MySQL Facebook contest application with forms, invitations and AJAX frontend interactions.
Selected work
LLM Fine-Tuning & Local Inference
Qwen2.5-Coder 0.5B logo model with LoRA experiments, a custom PyTorch training loop, GGUF quantization and CPU inference.
Recorded CPU benchmark: 24/24 valid outputs, 21/23 brand names preserved, 26.2 s mean generation time. Small synthetic-data experiment.
GitHub
Results
Multilingual Career Model
Local model trained on CV and career Q&A, with translation before and after inference, output verification, held-out evaluation and hallucination checks.
Recorded multilingual regression: 42/54 (77.8%) against a 90% gate. Not promoted; fine-tuning experiment without RAG.
GitHub
Evaluation protocol
AI-Assisted Development
Claude Code workflow with Jira tasks, isolated Git worktrees, per-task environments, browser verification, a UAT gate, layered instructions and code-graph support.
Public guide and templates for verification gates and seven-language AI-text quality checks; environment-specific ws CLI is not included.
GitHub
UAT workflow
AI Benchmark & Market Intelligence
Collection and normalization pipelines turn model accuracy, model pricing and cloud GPU pricing data into comparison views.
15 benchmark sources · 12 GPU pricing sources · 3 model-pricing sources.
Resilient Web Data Pipelines
Bright Data and direct HTTP/cURL scraping workflows with retries, honeypot detection and faulty-data checks before downstream use.
500+ URLs processed per month.
Multilingual Content & Reviews
Translation infrastructure for articles and static content, multilingual AI-text quality checks, and review analysis for spam and human-review prioritization.
500+ articles and static content across seven languages.
Private Codebase RAG
Fully local Ollama, DeepSeek-R1 and ChromaDB setup for querying PHP code and database schemas.
50,000+ lines of PHP · about 40% less manual inspection time.
SüperLig Data Explorer
Live Next.js application with source-linked match records, season and team filters, and an explicit decision-impact scoring model.
Three seasons · 1,028 match records · 4,754 card events. Suspension and VAR research remains incomplete; scores are not evidence of referee bias.
Live demo
GitHub
Methodology
PHP Coding-Agent Benchmark
Task-based evaluator for PHP coding agents covering snapshot isolation, idempotency, persistence, concurrency and crash recovery, with static analysis and mutation scoring.
Five task variants with recorded runs and calibration controls. Small run counts limit comparisons between closely scoring models.
GitHub
Results
Download CV