Adamus S.A. · Szczecin
Architected and deployed an enterprise multi-agent system on Azure, incorporating prompt engineering to automate the quotation process.
>
AI Engineer specializing in AI Agents, Prompt Engineering and Azure Solution Architecture - designing secure, scalable agentic workflows and enterprise systems that bridge AI models with robust Azure architecture and system security.
01 - who I am
I build agentic AI systems that turn language models into dependable enterprise tools. My work spans the full stack of modern AI engineering - from prompt engineering and agent orchestration with LangChain & LangGraph, to secure Azure architectures and MLOps pipelines that keep models measurable and production-ready.
Previously a software and robotics engineer, I bring a pragmatic, systems-first mindset: clean APIs, reproducible experiments, and security baked in from the start.
Designing multi-agent systems with clear boundaries, structured outputs and human-in-the-loop safety.
Deploying on Azure with architecture focused on scalability, observability and security.
PyTorch + MLflow experimentation, reproducible training and measurable evaluation.
02 - toolbox
03 - career path
Architected and deployed an enterprise multi-agent system on Azure, incorporating prompt engineering to automate the quotation process.
Designed scalable backend architecture and APIs for spatial and simulation platforms using Docker and PostgreSQL.
Developed backend services and REST APIs for client projects as an independent freelancer, working with relational databases.
Programmed a collaborative robot (cobot) to synchronize movements with human motion captured via computer vision.
04 - selected work
★ Featured project
Advanced Reinforcement Learning (SAC / PPO / A2C) for autonomous Robot Sumo combat - featuring competitive self-play in continuous action spaces.
Chrome extension that enhances the Gemini web experience.
Quadcopter control system simulation based on linear controllers with extensive visualization.
High-fidelity simulation of rocket flight dynamics and control - ideal for ML/RL training and GNC research.
Multiparadigm DQN + Imitation Learning project focused on solving a 2×2 Rubik’s cube.
ML model for collision detection trained on accelerometer sensor data with PyTorch.
RAG-based application that answers questions and generates quizzes & flashcards from provided materials.
05 - background
Poznan University of Technology
Department of Automation, Robotics and Electrical Engineering
Poznan University of Technology
Department of Computer Science
06 - contact
Open to AI engineering roles, agentic-system projects and Azure architecture challenges. Drop me a line - I usually reply within 24 hours.