Technical University of Munich
Oct. 2024 – Present
M.Sc. Informatics
Munich, DE
Leo Eckert
Master's student in Informatics at TUM, specializing in computer vision and medical imaging, with experience building healthcare and machine learning systems.
01Education
UC San Diego
Sept. – Dec. 2025
Quarter abroad
San Diego, USA
Technical University of Munich
Oct. 2020 – Sept. 2024
B.Sc. Informatics, App. field Medicine
Munich, DE
Umeå Universitet
Aug. 2022 – Jan. 2023
Semester abroad
Umeå, SE
02Experience
Machine Learning Engineer, Working Student
Feb. 2026 – Present
Avelios Medical GmbH
Munich, Germany
- Developing an automated medical speech-to-text structuring system using LLMs, speech recognition models, and gRPC
Backend Engineer, Working Student
Oct. 2023 – Sept. 2024
Avelios Medical GmbH
Munich, Germany
- Contributed to the development of a large scale, modern Hospital Information System consisting of micro-services, based on Spring Boot and gRPC
- Developed integration methods and processes to communicate with legacy systems in use, helping hospitals integrate their complex legacy systems into modern operations, using HL7 and FHIR standards to exchange medical data
03Research Experience
Dehazing Echocardiography Images
May – Sept. 2025
Python · PyTorch
- Developed a CycleGAN-based model to dehaze noisy echocardiography images in difficult-to-image patients, improving clarity and preserving key cardiac structures for downstream analysis
- Presented the work at the MICCAI 2025 conference in South Korea
- Achieved 2nd place in the Dehazing Echo 2025 Challenge organized by Fudan University
Uncertainty Estimation for Gaussian Splatting
Aug. – Nov. 2024
Python · PyTorch
- Bachelor's thesis (Grade: 1.0 — German system, equivalent to A+) on estimating reconstruction uncertainty in 3D Gaussian Splatting (3DGS) scenes, addressing challenges like viewpoint diversity and occlusions
- Used a spatial perturbation field and Bayesian approximation to assess uncertainty in pre-trained 3DGS scenes without modifying the training pipeline
- Created a spatial uncertainty field visualizable through an additional color channel, enhancing interpretability
04Projects
Memory Lanes — HackaTUM
Nov. 2024
React · TailwindCSS · Whisper
- Won the Popular Vote Award among 211 projects and placed 2nd in the City of Munich challenge with a service connecting users to personal stories tied to city locations
- Integrated Google Maps, QR code-based storytelling, and audio playback with live multilingual subtitles
05Skills
Programming
Python, Java, C, C++, JavaScript, HTML/CSS
Technologies
Docker, gRPC, Spring Boot, PyTorch, OpenGL, Git, React
Languages
German (native), English (C2), Swedish (A2)
06Interests
- Digital & Analog Photography
- Swimming
- Running
- Playing Trombone