Bachelor's degree

Built and shipped an AI-powered job application platform from zero to 200+ users with $0 marketing budget.

→ Chrome Extension (MV3) supporting 8+ job platforms (LinkedIn, Indeed, Glassdoor, JustJoin.it, Pracuj.pl) with unified platform abstraction layer.

→ Privacy-first architecture: CV parsing runs locally in browser via Offscreen Document API (PDF.js, Mammoth)

→ Multi-model LLM pipeline with primary/fallback chain, structured JSON validation, 12-language support, HMAC-secured extension auth

→ Remote config system with TTL caching for platform selectors, zero-downtime updates without extension releases

→ Decent Google ranking for brand keywords through technical SEO, $0 ad spend

Stack: Next.js 15, TypeScript, Chrome Extension MV3, Supabase, Stripe, Tailwind CSS

Creator / Product Owner

→ Architected a multi-agent AI platform: 8 specialist agents with autonomous handoffs, 3-tier RAG pipeline (general + domain-specific + offers knowledge bases), 768-dim vector embeddings, intent classification, all on Cloudflare edge

→ Built interactive geospatial UI: Leaflet.js marker clustering, isochrone visualization (walk/drive/bike zones) triggered by natural language through custom agent token protocol, polygon filtering with ray-casting algorithm

→ Delivered voice-powered field operations SaaS: Whisper transcription (large-v3, int8 quantization, ~15s on CPU), WhatsApp bot integration, AI-powered task classification, Kanban board with drag-and-drop, Stripe subscriptions, live in production with paying customers

→ Shipped human design SaaS: dynamic SVG bodygraph renderer, guest-to-user conversion with cookie tracking, admin CMS with 202 database-driven content fields, Stripe billing

→ Full TypeScript stack: Next.js 15, Cloudflare Workers/D1/KV/R2/Vectorize, Stripe, Docker, Ra

Full Stack Engineer

Built desktop image processing toolkit for numismatic auction house, handling 2,000+ coin images per auction cycle.

→ 5-tool pipeline: auto-crop (AI coin detection + threshold modes), batch renaming, image pair merging, auto-level adjustment, AI background removal (BiRefNet/ONNX with CUDA GPU acceleration)

→ Reduced manual editing time by 90%, replaced DAYS of Photoshop work with one-click batch processing, to ~1 hour.

→ Built with Python, customtkinter, PIL, NumPy, ONNX Runtime, PyTorch

→ Parallel processing, ICC profile preservation, recursive folder handling

Image Automation Specialist

full stack dev

Know about Ömer Uğur's work experience, education, bootcamps, certifications, and much more. Ömer Uğur