Skip to content
View frissonitte's full-sized avatar
💭
Feel free to reach out for collabs & questions!
💭
Feel free to reach out for collabs & questions!

Highlights

  • Pro

Block or report frissonitte

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.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. 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
frissonitte/README.md

Hi, I'm Emirhan 👋

Applied ML Engineer | Computer Vision · LLM Systems · Production ML

I specialize in bridging the gap between AI prototypes and production. Whether it's training robust Computer Vision models, designing Retrieval-Augmented Generation (RAG) pipelines, or building high-throughput streaming architectures, my focus is on engineering scalable, real-world ML systems.


🚀 What I Do

  • Applied AI & ML: Training deep learning (CV) and generative AI (LLM/RAG) models, deploying them as production-ready microservices with a focus on out-of-distribution (OOD) generalization and explainability.
  • MLOps & Data Pipelines: Building automated training pipelines, CI/CD for model deployment (GitHub Actions, Docker), and out-of-core streaming architectures (Dask, drift detection).
  • Backend & Systems Engineering: Architecting fault-tolerant backend workflows, high-concurrency data ingestion pipelines, and resilient multi-threaded systems.

🛠 Core Stack

  • ML & Deep Learning: PyTorch TensorFlow Scikit-Learn Transformers Sentence Transformers
  • Computer Vision & NLP: OpenCV Image Processing RAG ChromaDB Spotify Annoy
  • Data & Backend: Python FastAPI Flask PostgreSQL Dask (ETL) Pandas SQL
  • DevOps & Infrastructure: Docker GitHub Actions (CI/CD) Git Linux

🔬 Featured Projects

Project Description
wbc-analyzer End-to-end WBC classification system deployed as a Flask REST API. Features a custom architecture (DenseNet121 + WBCAttention + MedSwish). Incorporates inference-time domain adaptation achieving an 89.05% out-of-distribution (OOD) accuracy (+32.09 pp boost). Includes a multi-modal LLM agent (GPT-4o & Gemini) for clinical XAI insights. Preprint published on ResearchGate.
rag-project-assistant Hybrid-source RAG system answering questions about my portfolio projects. Combines AST-extracted code structure with curated documentation, using similarity-threshold gating (ChromaDB, L2 < 1.40) to prevent hallucination on out-of-scope queries. Deployed on Hugging Face Spaces (Docker, FastAPI, Groq Llama 3.3 70B) with rate-limited public API and a live chat widget.
kinematic-action-recognition Full end-to-end ML pipeline on 10 GB motion-capture sensor data. Features out-of-core ingestion with Dask, real-time streaming with drift detection (81 windows/sec), and a LightGBM/RandomForest ensemble achieving 0.94169 accuracy on Kaggle.
popcorn-wagon Hybrid movie recommender engine built with Dask/Pandas for scalable ETL and Spotify Annoy for sub-millisecond similarity search. Integrates collaborative filtering (SVD) and content-based filtering.

Pinned Loading

  1. wbc-analyzer wbc-analyzer Public

    AI-powered pathology assistant for White Blood Cell classification, featuring a custom lightweight DenseNet121 architecture (WBCAttention + MedSwish), inference-time domain adaptation, Flask REST A…

    Python

  2. kinematic-action-recognition kinematic-action-recognition Public

    A full end-to-end pipeline for classifying factory operator actions from high-frequency motion-capture sensor data. Covers out-of-core data engineering, time-series feature extraction, unsupervised…

    Python

  3. popcorn-wagon popcorn-wagon Public

    A Flask-based movie recommendation website.

    Python

  4. rag-project-assistant rag-project-assistant Public

    A RAG (Retrieval-Augmented Generation) chatbot for querying personal project documentation. Built to replace hallucination-prone LLM responses with grounded answers extracted from actual project so…

    Python