Hi there! I'm Andrey Korchemkin

Welcome to my GitHub!

I'm a ML Engineer specializing in production AI systems:

• Vision-Language Models (VLM) & Large Language Models deployment
• Computer Vision & Document AI (OCR, layout analysis)
• MLOps & Infrastructure (Docker, vLLM, multi-GPU inference)
• NLP & Structured Data Extraction

Professional Experience

ML Engineer @ SibinTek Soft (Rosneft subsidiary)

• Built AI verification system for construction documentation using Qwen VL, reducing manual verification from 5-6 days to 2-3 hours
• Developed multi-page PDF->Excel digitization pipeline (KS-2 forms) with Qwen VLM + matching algorithms
• Deployed production VLM inference on 3x RTX A4000 cluster via vLLM + Docker, processing 800 pages/hour with parallel optimization
• Created AI assistant prototype for Excel with natural language formula generation

Education & Research

Bachelor's in Software Development - Diploma project on railway safety systems using deep learning on point clouds (Open3D) and RGB data.

Current Focus

• Vision-Language Models for document understanding & verification
• Production LLM deployment optimization (quantization, inference tuning)
• Multi-agent AI systems with Model Context Protocol (MCP)

Tech Stack

Core: Python, PyTorch, Transformers, vLLM, LangChain
CV/VLM: OpenCV, Open3D, Qwen VL, OCR pipelines
MLOps: Docker, CUDA, Linux, multi-GPU inference, A4000/H100
Tools: Scikit-learn, HuggingFace, Streamlit, Gradio, Flask
Also: SQL, Java, C++, RabbitMQ, JavaScript