Looking for a job

Petros Vezalis

@petrosvezalis

Message

I build traceable RAG assistants and retrieval systems for industrial technicians.

Greece

Message

What I'm looking for

I'm looking to build and improve production AI systems, especially RAG and LLM applications, where I can work on retrieval quality, grounding, evaluation, and practical data pipelines.

At Atlantis Engineering, I built traceable answers for a production LLM assistant used by industrial technicians, linking generated claims to source documents, pages, and sentences. I also extended the assistant with schema-aware text-to-SQL for numeric and time-sensitive questions.

I've improved retrieval relevance through structure-aware chunking, PDF parsing, pgvector indexing, query rewriting, and re-ranking. My thesis examined RAG optimization for an industrial LLM assistant, and I also built a fraud-detection system using machine learning classifiers and feature engineering.

Experience

Work history, roles, and key accomplishments

Junior AI Engineer

Atlantis Engineering

Sep 2025 - Oct 2025 (1 month)

Made answers traceable in a production LLM assistant serving industrial technicians, building a citation pipeline that ties each generated claim to its source document, page and sentence. Extended the assistant past unstructured manuals with schema-aware text-to-SQL, routing numeric and time-sensitive questions to execute SQL instead of vector search.

LLM RAG Citations Text To SQL Python SQL

AI Engineer Intern

Atlantis Engineering

Jun 2025 - Aug 2025 (2 months)

Raised the assistant's contextual relevance from 0.69 to 0.79 with a custom structure-aware + semantic chunking strategy, chosen by benchmarking 6 techniques on an expert-annotated Q&A set. Cut ingestion noise across scanned, multilingual PDFs via reading-order extraction (pdfplumber), table detection (Camelot) and normalization, indexing prose and tables separately in PostgreSQL/pgvector. Improve

RAG Chunking Pdfplumber Camelot PostgreSQL Pgvector OpenAI Ranking

Education

Degrees, certifications, and relevant coursework

University of Macedonia

Bachelor of Science, Computer Science and Technology

Grade: 7.6/10

Activities and societies: Relevant coursework: Machine Learning, Neural Networks, Big-Data, Artificial Intelligence, Cloud Computing, Parallel and Distributed Computation, Database Systems

Bachelor of Computer Science and Technology with a minor in Artificial Intelligence, graduated in September 2025 with a GPA of 7.6/10.