Liul T.

I build AI systems that move from idea to production. Top Rated on Upwork with a 100% Job Success Score and $10K+ earned, I help businesses turn data, documents, images, and business processes into reliable AI systems. When a project involves extracting value from data, automating a knowledge workflow, understanding images, or integrating AI into an existing product, I can take it from data preparation and system design through model development, evaluation, API integration and deployment. My core work includes: AI/ML and Data Science I have experience across the full machine learning pipeline, including data collection and cleaning, exploratory analysis, feature engineering, preprocessing, model development, training, evaluation, optimization, inference, API development, containerization, and deployment. My main tools include Python, SQL, Scikit-learn, TensorFlow, PyTorch, FastAPI, Flask, Docker, PostgreSQL, and pgvector. Agentic AI and RAG I build document processing and agentic RAG systems for real-world business use cases. My experience includes document ingestion, OCR and VLM-based extraction, retrieval, structured data processing, multi-step agent workflows, tool calling, SQL execution, citations, authentication, and production deployment. I work with LangGraph, LangChain, CrewAI, OpenAI SDK, MCP/FastMCP, vector databases, and modern LLM systems. Computer Vision I have worked continuously on computer vision projects for more than a year, including image preprocessing, defect detection, model training, inference pipelines, optimization, and production deployment. Through my work with Elunic AG, a Germany-based software company, I have worked on industrial manufacturing computer vision projects and contributed to defect detection solutions for 20+ companies. Selected projects DocMind: Enterprise Agentic RAG & Document Intelligence Platform Architected and deployed a document intelligence platform that transforms complex PDFs into structured, queryable data. Containerized and deployed to Oracle Cloud. The Ledger: Enterprise Agentic Event Store & Audit Infrastructure Built an event sourcing and audit infrastructure for autonomous AI agents using Python, FastMCP, LangGraph, and PostgreSQL. AI-based Tuberculosis Screening Using Cough Sounds Developed a CNN-based cough-sound classification system for tuberculosis screening, trained on nearly 10,000 recordings, achieving 82% accuracy and 0.90 ROC-AUC. I enjoy taking insights from data and turning them into useful systems that solve real problems. I am especially interested in AI/ML, Data Science, RAG, Agentic AI, and Computer Vision, with a strong focus on building deeper expertise in agentic systems, RAG, and computer vision. I am naturally curious about new technologies and comfortable working in unfamiliar technical areas. Whether the project involves an established ML workflow or a new AI problem, I focus on understanding the real requirement, building the right solution, and getting it into a form that can actually be used.