Machine Learning Engineer | Thessaloniki, Greece

📧 constantinos.giantsios@gmail.com | LinkedIn | GitHub

Machine Learning Engineer with 5+ years of expertise specialising in NLP, Recommendation Solutions, Search, and Agentic AI. Proven track record in designing, deploying, and scaling production-grade machine learning models, complex AI architectures, and real-time user profiling platforms.

Aristotle University of Thessaloniki (AUTh), Thessaloniki, GR — 09/2020 – 06/2022 M.Sc. in Data and Web Science, GPA: 9.5/10.0 Thesis: Interpretable Multi-Label Learning in Biomedical Texts, Grade: 9.5/10.0

University of Macedonia (UoM), Thessaloniki, GR — 09/2016 – 09/2020 B.Sc. in Applied Informatics, GPA: 8.4/10.0 (top 9%) Thesis: Detecting toxic comments and minimising unintentional prejudice using neural networks, Grade: 10.0/10.0

Kaizen Gaming, Machine Learning Engineer, Thessaloniki, GR — 05/2025 – Present

• Engineered robust data pipelines employing Apache Spark/PySpark and Databricks on Microsoft Azure, processing large-scale interaction data to power an agentic AI Chatbot for customer experience analysis.
• Co-designed and implemented a multi-agent sports betting Copilot using Python, LangChain, and LangGraph, delivering an interactive assistant capable of complex reasoning and resolving nuanced user queries.
• Deployed complex agentic AI solutions as scalable RESTful API services utilising FastAPI, Docker, and Redis, ensuring high availability and low-latency production responses.
• Established comprehensive evaluation frameworks and dashboards leveraging LangSmith and RAGAS, enabling continuous monitoring, rapid debugging, and improved reliability of LLM applications.
• Tech stack: Python, Apache Spark/PySpark, Hugging Face, Neo4j, Docker, FastAPI, Microsoft Azure, Redis, Databricks, LangChain, LangGraph, LangSmith

Independent Consultant, Machine Learning/Data Engineer, Thessaloniki, GR — 04/2023 – 09/2023

• Formulated a specialised RAG architecture using Elasticsearch, enabling highly accurate, hybrid search of influencer profiles for targeted brand campaigns.
• Created a scalable pipeline leveraging OpenAI APIs and Hugging Face models, streamlining the workflow for generating domain-specific educational materials.
• Designed an automated, end-to-end system for generating professional headshots, harnessing PyTorch and OpenCV, efficiently deployed via FastAPI and Docker.
• Tech stack: Python, PyTorch, Apache Spark/PySpark, Hugging Face, OpenAI, Elasticsearch, Docker, FastAPI, OpenCV, Pinecone

Atypon Systems (Wiley), Machine Learning Engineer, Thessaloniki, GR — 06/2020 – 05/2025

• Architected a hybrid retrieval system (Elasticsearch) and integrated LLMs (Gemini, Llama) via LangChain and Self-RAG. Fine-tuned bi-encoders (MiniLM), boosting recommended article CTR by 100% and search CTR by 20%.
• Developed a comprehensive taxonomy and automated tagging system using deep learning Transformers, leading to a 30% increase in Success Search Rate (SSR) and reducing manual tagging time by over 90%.
• Orchestrated high-performance model serving utilising NVIDIA Triton Inference Server and optimised document processing pipelines with quantization and ONNX to ensure scalable, low-latency production inference.
• Implemented a real-time user profiling system from streaming event data to generate personalised promotions, driving a 300% increase in new user registrations and a 200% increase in authenticated sessions.
• Built testing pipelines to evaluate embedding retrieval (SciRepEval, BEIR) and continuously measure LLM faithfulness and answer relevancy using the RAGAS framework and LLM-as-a-judge techniques.
• Tech stack: Python, Java, PyTorch, TensorFlow 2.0, Apache Spark/PySpark, Apache Beam, Hugging Face, PostgreSQL, MongoDB, MySQL, Elasticsearch, MLflow, ONNX, Docker, Flask, FastAPI, Apache JMeter, GCP, NVIDIA Triton Inference Server, Redis, PubSub, Dataflow, LangChain, LangGraph, LlamaIndex

Programming Languages: Python, Java, Scala

Tools/Frameworks: TensorFlow 2.0, PyTorch, Apache Spark/PySpark, Apache Beam, scikit-learn, Hugging Face, spaCy, MLflow, ONNX, OpenCV, NVIDIA Triton Inference Server, PubSub, Dataflow, LangChain, LangGraph, LlamaIndex, LangSmith, Langfuse

MLOps & Engineering: REST, Microservices, Git, CI/CD, Maven, Docker/Compose, Apache JMeter, Flask, FastAPI

Cloud & Data: Google Cloud Platform (GCP), Microsoft Azure, Elasticsearch, Neo4j, Redis, PostgreSQL, Pinecone (Vector DB), MongoDB, Databricks

AI & ML: Classical ML (Regression, Clustering, SVMs), Deep Learning (LLMs, Transformers, CNNs), Dimensionality Reduction, Prompt Engineering, Agentic AI

• Google Hash Code: Achieved 2nd place among 12 University of Macedonia teams and ranked in the top 20% overall in Greece.
• Kaggle Competition: Ranked in the top 6% worldwide in the 'Jigsaw Unintended Bias in Toxicity Classification' competition.
• Agentic AI MOOC (Legendary Tier) – UC Berkeley Center for Responsible, Decentralized Intelligence, Fall 2025.
• Machine Learning in Production – DeepLearning.AI via Coursera, May 2023.
• Deep Learning Specialization (5 Courses) – DeepLearning.AI via Coursera, Oct 2020.
• Machine Learning – Stanford University via Coursera, Nov 2019.

Greek (Native), English (C2/Proficient)