Personal portfolio / 2026

Thessaloniki, Greece

NikolaosKaramousalidis

Full-Stack AI Engineer

From research questions to working AI systems.

Fig. 01 / Connected intelligence
Information → understanding Systems & research
Research, systems, and the engineering in between.Explore the portfolio

01 / The portfolio

About

I’m a Full-Stack AI Engineer in Thessaloniki, Greece, working across AI systems and web applications. My work spans RAG, semantic search, FastAPI microservices, and LLM evaluation. I also publish research on LLM-generated logic programs.

01Retrieval-augmented generation02Semantic search03LLM evaluation

02 / The portfolio

Experience

Current role

Full-Stack AI Engineer

ARI Foods S.A.

  • Developed and evaluated an AI-assisted label-review prototype comparing PDF artwork with Excel product specifications, with evidence-linked findings for human review.
  • Built HR attendance and daily staff-allocation dashboards integrating ERP data, role-based access, live TV displays and automated Excel exports.
  • Improved refrigeration-alert automation using Microsoft Graph, Teams and Azure, including acknowledgments, reminders, notification retries and operational health checks.
PythonFastAPILLMsSQL ServerDockerMicrosoft GraphAzure
Remote

AI Engineer

Gaddr

  • Built a FastAPI semantic search microservice for user profile discovery using Hugging Face embeddings, PostgreSQL and pgvector.
PythonFastAPIHugging FacePostgreSQLpgvector
Thessaloniki, Greece

Full-Stack Developer

Municipality of Thessaloniki

  • Developed and deployed 10+ new frontend features for PotiZo, a web platform used by the Municipality of Thessaloniki.
  • Collaborated within a multinational student team to develop features, resolve software bugs and optimise the user experience.
ReactTailwind CSSREST APIs

03 / The portfolio

Projects

Selected work in LLM evaluation, document retrieval and applied machine learning.

Project / 02Applied AI
Workflow overview
  1. 01Upload
  2. 02Index
  3. 03Retrieve
  4. 04Answer

Retrieval-Augmented PDF Question Answering

A RAG application that lets you upload several PDFs and query them in natural language.

  • Document parsing, chunking, embeddings and semantic retrieval built with LangChain and FAISS.
  • Conversational history so follow-up questions stay context-aware.
PythonLangChainOpenAIFAISSRAG
Project / 03Machine learning

Heart Disease Risk Classification & Cluster Analysis

Classification and clustering models trained on a clinical dataset of 295 samples.

  • Random Forest classification alongside K-Means cluster analysis.
  • 87% accuracy, 75% disease recall and a 0.80 F1-score through hyperparameter tuning and class balancing.
PythonScikit-learnRandom ForestK-Means

04 / The portfolio

Publications

Peer-reviewed research on large language models and logic programming.

01

A Multi-Agent Benchmarking Framework for Evaluating the Performance of Large Language Models in Logic Programming

Nikolaos Karamousalidis, Petros Kefalas

SN Computer Science · Jul 30, 2026

This journal article extends the paper presented at the 2025 Balkan Conference in Informatics, with expanded experiments, benchmarking methodology, and analysis.

  • Broadened the 12-task benchmark to cover constraint logic programming, declarative pattern matching, and grid-based state-space problems
  • Expanded the experiments from five models to seven newer LLMs across code generation and evaluation
  • Added sandboxed SWI-Prolog execution and structural/runtime metrics, including predicate use, logical inferences, and memory
  • Added a multi-evaluator reliability study using Fleiss’ kappa, finding only slight overall agreement (κ = 0.1716)
02

Performance of Large Language Models in Automatic Generation of Logic Programs

Nikolaos Karamousalidis, Petros Kefalas, Ioanna Stamatopoulou

11th Balkan Conference in Informatics, Tirana · Oct 2025

  • Experimental evaluation of LLMs for automatic Prolog code generation
  • Designed multi-agent systems for problem understanding, code generation, and evaluation
  • Developed custom logic-programming evaluation metrics

05 / The portfolio

Skills & technologies

The tools behind my AI systems, research and web applications.

25 technologies · 6 areas

06 / The portfolio

Background

Education

M.Sc., Artificial Intelligence and Data Science

University of York, Europe Campus

Expected graduation: June 2027

Relevant coursework: Big Data, Deep Learning, Machine Learning.

B.Sc., Computer Science

University of York, Europe Campus

Graduated July 2025

3.3 GPA. Relevant coursework: Cloud Computing, Web Programming, Software Engineering.

Certifications

Azure AI Fundamentals

Microsoft

Activities

Volunteer — “Acting Together for Nature”

Outward Bound Romania / European Solidarity Corps

August 2026

Two-week European Solidarity Corps environmental and community project in Romania, contributing to mountain trail rehabilitation, youth activities and the renovation of community facilities.

Community Kitchen Volunteer

Odyssea Youth Center

November 2025 – June 2026

Prepared and packaged vegetarian meal boxes and assisted with distribution to individuals in need as part of Odyssea Youth Center's community kitchen initiative.

Class Representative

University of York, Europe Campus

October 2025 – October 2026

Communicated student feedback to faculty, helped resolve academic issues and collaborated with professors.

07 / CV

Curriculum vitae

The full record of experience, research, and education.

08 / The portfolio

Contact

Have a question or an opportunity? Send a message.

Email delivery is being configured.