Skip to content
View SerhanTelatar's full-sized avatar

Block or report SerhanTelatar

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
SerhanTelatar/README.md

Serhan Emre Telatar

AI Engineer · Deep Learning & Computer Vision · 3D Vision × Generative AI

Email LinkedIn GitHub


About

I'm an AI Engineer working at the intersection of 3D computer vision and generative AI — from diffusion-based virtual try-on to multimodal fine-tuning pipelines running in production. Currently pursuing an MSc in Artificial Intelligence at Queen Mary University of London, supervised by Shanxin Yuan, while working as a Deep Learning & CV Engineer.

My background is T-shaped: deep in 3D vision and generative modelling, broad across the ML production stack — from CUDA-level optimization to shipping full applications end-to-end.

role:        AI Engineer — Deep Learning & Computer Vision
focus:       3D Computer Vision, Generative AI (Diffusion, VLMs)
currently:   MSc AI @ Queen Mary University of London
based_in:    London, UK / Istanbul, TR

Experience

Deep Learning & Computer Vision Engineer · Inovako (2026 – Present) Multimodal fine-tuning pipelines (Llama 3.2 Vision, Qwen 2.5-VL) for industrial material classification, LoRA + quantization for on-GPU deployment, video-QA over VLMs, and a fully local LangGraph-orchestrated deep-research RAG agent.

Machine Learning Engineer · BNP Paribas TEB (2024) End-to-end credit-default prediction pipeline over 100K+ loan applications; statistical feature analysis that improved early-warning detection and reduced bad-debt provisioning.


Featured Projects

MeshVTON

Multi-View Virtual Try-On via 3D Mesh Conditioning

Conditions a frozen 12B FLUX.1-Fill diffusion transformer on 3D-aware renderings via a lightweight control branch + LoRA (~112.5M trainable params, <1% of backbone). Full 3D pipeline: SMPL-X regression, garment draping, multi-view rendering.

Diffusion 3D Vision SMPL-X LoRA

Studel

AI Exam Simulator — Native iOS

SwiftUI (MVVM) app with a RAG pipeline on GCP (Vertex AI + pgvector) that grounds quiz generation in a user's own lecture material, using an agentic workflow to plan, draft, and validate each question.

SwiftUI RAG Vertex AI Agentic Workflows

Multi-Agent Supply Chain Negotiation

Reinforcement Learning

Multi-agent RL environment where manufacturer and retailer agents co-adapt over 40-period newsvendor negotiations, converging to equilibrium-like joint policies under wholesale, buyback, and revenue-sharing contracts.

Multi-Agent RL Q-Learning Game Theory

More on GitHub

Explore full implementations, ablations, and write-ups on my pinned repositories.

→ View all projects


Tech Stack

Computer Vision OpenCV PyTorch3D CLIP YOLO

3D & Generative Diffusion NeRF 3D Gaussian Splatting SMPL-X

ML / DL Frameworks PyTorch TensorFlow Hugging Face ONNX

LLM / Agentic LangChain LangGraph OpenAI API Qdrant

Languages & Infra Python C++ TypeScript Docker Kubernetes GCP AWS CUDA


GitHub Stats



Education

MSc Artificial Intelligence — Queen Mary University of London (2025 – 2026) BSc Computer Science and Engineering — Sabancı University (2020 – 2025)


Open to collaborating on 3D vision, generative modelling, and applied ML research.

Pinned Loading

  1. Deep-Reinforcement-Learning-for-the-World-s-Hardest-Game Deep-Reinforcement-Learning-for-the-World-s-Hardest-Game Public

    Deep Q-Learning agent for a browser-based World’s Hardest Game clone. Play manually or watch your AI agent learn to beat the game live!

    JavaScript

  2. Supply-Chain-Decision-Analysis Supply-Chain-Decision-Analysis Public

    ENS491/492 - Developing Behavioral Models to Explain Supply Chain Decisions

    Jupyter Notebook

  3. AI-Tweet-Analysis-Automaton AI-Tweet-Analysis-Automaton Public

    JavaScript

  4. Trash-Classifier Trash-Classifier Public

    Jupyter Notebook

  5. plate_recognition plate_recognition Public

    Jupyter Notebook

  6. MeshVTON MeshVTON Public

    Jupyter Notebook