Ersan Erdem

I craft production grade AI image pipelines for photoreal portraits, creative styles, and explainable imaging. My work combines Stable Diffusion SDXL, LoRA, ControlNet, IP Adapter, and FaceSwap/InSwapper to deliver consistent identity, natural skin, and professional quality outputs. Recent highlights include: Deploying AI generators on Replicate with dynamic LoRA loading, QA scoring, and Docker CI/CD. Delivering studio quality headshots with ArcFace ID consistency and 4K tiled refinement. Creating anime/creative styles with LoRA + identity preserving FaceSwap pipelines, optimized for e commerce, HR, and content creators. Beyond AI, I bring an academic foundation as an Applied Mathematics MSc with a published paper “Web Application for Step Size Strategies” and an ICMME2021 conference presentation COVID 19 analysis with LaTeX Beamer. This background allows me to also support clients with LaTeX based technical documentation—ensuring accuracy in formulas, references, and structure. Tech stack: Stable Diffusion SDXL, LoRA, ControlNet, IP Adapter, ArcFace, InsightFace, FaceSwap/InSwapper, Gradio, Replicate/Cog, Docker, Python, LaTeX/Overleaf, Git/GitHub.