Bálint Mészáros
AI Engineer with Focus on Multi-Agentic Systems
Passionate AI Engineer specializing in Large Language Models, Computer Vision, and Edge AI. With hands-on experience at BMW AG and academic excellence from TU Munich, I develop scalable GenAI solutions and optimize models for real-world deployment.
Work Experience
Master's Thesis
BMW AG
- •Successfully developed an LLM-powered computer vision framework for in-vehicle scene understanding.
- •Established a novel benchmark for performance evaluation.
- •Fine-tuned model outperformed GPT-4o by over 20%.
Research Internship
TU Munich - Chair of Data Processing
- •Researched Vision Transformer models for efficient real-time image based blood cell classification on edge devices.
- •Built a highly efficient transformer-based architecture with PyTorch, achieving 87% accuracy while being almost 200 times smaller than DenseNet with only a 2.5% performance gap.
AI Software Developer Intern
BMW AG
- •Engineered deep learning and GenAI-powered car features using Hugging Face and TensorFlow for scalable solutions.
- •Analyzed and processed terabytes of customer data with Pandas, AWS Athena and S3; deployed models on cars and AWS EC2.
- •Coordinated work with external software suppliers to ensure code quality and seamless integration.
ML and Data Analytics Intern
MORESOPHY GmbH
- •Implemented an NLP pipeline for automated data crawling and preprocessing, storing results in Elasticsearch.
- •Efficiently handled millions of documents using Pandas and integrated predictive modeling with ktrain.
- •Set up dashboards for monitoring, automated workflows with ArgoWorkflows, and managed Dockerized pipeline on Kubernetes.
Software Tool Developer Intern
Robert Bosch Kft.
- •Designed SIL and MIL test frameworks in C and MATLAB for automotive ECU validation.
Education
M.Sc. in Electrical Engineering
TU Munich
Thesis
Scalable Object Detection in Car Interiors with Vision Foundation Models
Grade: 1,3
B.Sc. in Electrical Engineering
TU Budapest
Thesis
ECU Design for BME FUSE Formula Student Team's Electric Vehicle
Grade: 1,0
Skills & Expertise
Programming Languages
Tools & Frameworks
Technologies & Methods
Projects & Research
Local LLM Assistant
- •Local chatbot with long-term memory using SQLite and RAG.
- •Used LangChain, LangGraph, and Streamlit for implementation.
Applied Machine Intelligence
- •Constructed an end-to-end deep learning framework for blood cell segmentation and classification with PyTorch.
- •Deployed a Dockerized application on a Kubernetes cluster.
RoboCup@Home
- •Developed plane detection and object grasping algorithms for a Toyota HSR robot using Python/C++, integrated within ROS.
Applied Reinforcement Learning
- •Created a reinforcement learning agent in C++ trained with approximate Q-learning for the HaxBall game.
About Me
In My Free Time
I love running, traveling, hiking with friends, and capturing urban and nature photography. I enjoy meeting new people and socializing. In summer, I play volleyball, frisbee, and enjoy barbecues in the garden.