Andrii Tynskyi

Junior AI & Software Engineer

Applied ML & GenAI / Python / RAG / scikit-learn / REST API

About

I'm a Computer Science student at ZUT in Szczecin, building a foundation for junior AI Engineer, Machine Learning Engineer, or software engineering roles. My recent work explores applied ML and GenAI through reproducible Python projects: document-grounded RAG, LLM response evaluation, and a scikit-learn model served through an API.

I bring broader software engineering experience from university projects and an internship at SOFTSERVICE TECHNOLOGIES: Java and Spring Boot APIs, a PHP MVC application, SQL data modelling, Android apps with Firebase, IoT prototypes, unit testing, Docker, and technical documentation. These foundations help me build AI features as maintainable software, not isolated demos.

My current focus is Python, applied machine learning, LLM evaluation, and reliable AI application design.

Location Szczecin, Poland
University ZUT Szczecin, Computer Science
Languages Ukrainian (native), English (B2), Polish (B1/B2)
Open to Szczecin-based; open to local, hybrid, and remote roles

What I work with

AI & Machine Learning

Python RAG LLM tool calling Prompt grounding LLM evaluation scikit-learn pandas NumPy

Backend

FastAPI Java Spring Boot REST API HTTP/JSON Maven

Databases

MySQL MariaDB H2 JPA/Hibernate SQL

Mobile

Android Java XML layouts Firebase Android Studio

Tools & testing

Git Postman JUnit Mockito unittest GitHub Actions Docker VS Code IntelliJ

Also used

HTML/CSS C/C++ C# Python Arduino Unity

Mobile & embedded

Android Firebase Arduino STM32

Projects

DocuAgent — grounded document assistant

view on github →

Python, SQLite FTS5/BM25, RAG, LLM tool calling, GitHub Actions

A document assistant that retrieves Markdown evidence, passes relevant passages to an OpenAI-compatible model, and returns cited answers. Includes bounded tool use, a restricted calculator, offline tests, and a small retrieval benchmark.

Measured on the included synthetic corpus: 19/19 expected sections retrieved in the top 3; this is a small retrieval benchmark, not a factuality guarantee.

LLM Eval & Security Lab

view on github →

Python, RAG evaluation, citation checks, prompt injection, CI

An offline evaluation harness for saved LLM answers. Checks citations against supplied sources, expected refusals, and basic prompt-injection resistance, then exports a JSON report for CI. Its lexical grounding check is explicitly a heuristic, not a semantic fact checker.

Customer Churn ML Platform

view on github →

Python, pandas, scikit-learn, FastAPI, reproducible evaluation

A complete tabular ML workflow: schema validation, preprocessing pipeline, stratified holdout evaluation, model persistence, and a FastAPI prediction endpoint. On its 200-row synthetic holdout, the sample model reached ROC-AUC 0.744 and recall 0.819; these are demonstration results on generated data, not real customer outcomes.

PHP 8, custom MVC framework, MySQL, Docker, LESS CSS

Movie catalog web app — browse films, search by title/director, filter by genre and streaming platform, leave reviews with ratings, save favorites. Has a full admin panel for managing the catalog. I built the MVC framework from scratch (routing, templating, ActiveRecord ORM) to understand how frameworks like Laravel actually work inside.

Dockerized MySQL setup, 7 migration files, three CSS themes (dark/light/contrast), and AJAX-powered favorite toggling.

Multikino REST API

view on github →

Spring Boot, Java, JPA, H2/MySQL, JUnit, Postman

Backend for a cinema ticket booking system. You can add movies, schedule screenings, register users and make reservations. The API checks if there are enough seats before confirming a booking and gives them back when you cancel. I wrote it to practice building a proper layered Spring Boot app with actual business logic, not just CRUD.

Covers 20+ endpoints, has 7 JUnit tests with Mockito, and the README has example Postman requests for everything.

Android, Java, Firebase, XML

Gym management app — tracks subscriptions, payments, and workout stats. Connected to Firebase for cloud storage. I also wrote the full technical docs for this one.

SIYS (Start Improving Yourself)

Android, Java, XML

Self-improvement app with goal planning, reminders, and relaxation features. I presented it in front of an audience and focused on making the UX feel natural and not annoying to use daily.

C++, PlatformIO, BT/NFC/RF/WiFi, PCB design

Hackathon project from PUTCyberDays — built an IoT device that scans Bluetooth, NFC, RF, and WiFi signals. Has a web server, keypad input, and display output. Designed the PCB schematic too.

Smart Home Lighting

Arduino, sensors, mobile app

IoT project — motion, temperature, and light sensors hooked up to Arduino with remote control through a phone app. Built a working prototype and demoed it live.

Lazior Rover

Fusion 360, STM32, 3D printing

Team project — we built a small rover. I handled the 3D models in Fusion 360, prepared files for printing, and helped with wiring and motor tests on STM32 Nucleo.

Zombie Apocalypse

Unity, C#, Blender

Survival game made with a team. I was the PR manager and also wrote some of the game logic. Handled communication between the team and put together the final presentation.

Background

2023 — now

ZUT Szczecin

B.Eng. Computer Science — Information Systems Engineering

Sep — Dec 2025

SOFTSERVICE TECHNOLOGIES LTD

Dev support intern — worked in a team, wrote documentation, handled tasks in a real dev environment.

2023

Secondary School No. 91, Lviv

Graduated with honors (Gold Medal)

Courses

Extra stuff

Start IT Academy, Samsung IT School, LeetCode, W3Schools

Contact

Looking for junior roles or internships. If you have something — hit me up.