About Me
I am Philipp, a machine learning engineer who recently finished his master's degree in that area. I have been studying machine learning since 2017 and have been working in that area ever since.
My main domain interests are in:
- • Natural language processing
- • Reinforcement Learning
- • Computer Vision
- • Robotics
I have taken many classes on more specialized domains including energy grids, climate science, material science, and finance. I am especially interested in latent diffusion models due to my work in my master's thesis and its applications in language modeling.
I have been working for over 7 years as a machine learning engineer training dozens of custom models which are running in many major hospitals as well as industry players and that led to a couple of papers.
Currently I am working at the Zuercher Kantonalbank creating an AI platform and writing analytical services.
Short CV
2017
2018
2019
2020
2022
2024
Professional Work
Asset Management Platform Engineer at Zuercher Kantonalbank (2024-Present)
Technologies:
Python, C#, JavaScript
Development of analytical services for asset management software. Designed and implemented an AI platform enabling chat-based interaction with MCP servers, RAG collections, and workflow creation. Integration of AzureOpenAI and self-hosted models (DeepSeek, Gemma3, Qwen3). Finetuning models using SFT for non-reasoning and GRPO for reasoning tasks.
Machine Learning Engineer at HS Analysis (2017-2024)
Technologies:
Python, PyTorch, TensorFlow, React, C#, WPF, C++, Docker, Kubernetes
Led development of the deep learning backend for data of any domain with a focus on computer vision. Custom model development for research and industry partners in various domains. Participated in multiple funded research projects and co-authored several peer-reviewed publications.
Fraunhofer Institute for Secure Information Technology (2019-2020)
Technologies:
Python, TensorFlow, PyTorch, C#, WPF, Selenium
Worked on models for NLP, focused on authorship verification. Contributed to large-scale hate speech detection project for social media for government agencies. Responsible for implementing research ideas and comparative analysis with existing methods.
Participation at Funded Research Projects
KI basierte Diagnostik des Lungenkarzinoms zur Unterstützung personalisierter Therapieentscheidungen
Funding: €400,000
Task: Development of the Deep Learning Model
While I was working at: HS Analysis GmbH
Project LinkHybridlösung mit kontaktloser VIso-TAktiler Diagnostik
Funding: €2,006,000
Task: Develop Segmentation Model for the Skin Detection and for the prediction of various parameters
While I was working at: HS Analysis GmbH
Project LinkhyPro – Integration hybrider Intelligenz in die Prozesssteuerung von Produktionsanlagen der Glasumformung
Funding: €680,000
Task: Development of the Deep Learning Model
While I was working at: HS Analysis GmbH
Project LinkAnalyse extremistischer Bestrebungen in sozialen Netzwerken (X-SONAR)
Funding: €3,100,000
Task: Development of a Model that can detect hate speech in social media posts
While I was working at: Fraunhofer Insitut für Sichere Informationstechnolgie
Project LinkPublished Papers
An Improved Topic Masking Technique for Authorship Analysis
Oren Halvani, Lukas Graner, Roey Regev, Philipp Marquardt • ARES • 2021
An Improved Topic Masking Technique for Authorship Analysis
Using Deep Learning to Distinguish Highly Malignant Uveal Melanoma from Benign Choroidal Nevi
Laura Hoffmann, Constance B. Runkel, Steffen Künzel, Payam Kabiri, Anne Rübsam, Theresa Bonaventura, Philipp Marquardt, Valentin Haas, Nathalie Biniaminov, Sergey Biniaminov, Antonia M. Joussen, Oliver Zeitz • Journal of Clinical Medicine • 2024
This study evaluates deep learning models for distinguishing highly malignant uveal melanoma from benign choroidal nevi based on fundus photographs.
Resemblance-Ranking Peptide Library to Screen for Binders to Antibodies on a Peptidomic Scale
Felix Jenne, Sergey Biniaminov, Nathalie Biniaminov, Philipp Marquardt, Clemens von Bojničić-Kninski, Roman Popov, Anja Seckinger, Dirk Hose, Alexander Nesterov-Mueller • International Journal of Molecular Sciences • 2021
Resemblance-Ranking Peptide Library to Screen for Binders to Antibodies on a Peptidomic Scale
Neuromorphic Vision mit Spiking Neural Networks zur Sturzerkennung im betreuten Wohnen
Sven Nitzsche, Brian Pachideh, Victor Pazmino, Norbert Link, Christoph Schauer, Lukas Theurer, Valentin Haas, Philipp Marquardt, Sergey Biniaminov, Jürgen Becker • IEEE Robotics and Automation Letters • 2021
Neuromorphic Vision mit Spiking Neural Networks zur Sturzerkennung im betreuten Wohnen
An Unsophisticated Neural Bots and Gender Profiling System
Oren Halvani and Philipp Marquardt • Conference and Labs of the Evaluation Forum • 2019
An Unsophisticated Neural Bots and Gender Profiling System
Nutzen der partizipatorischen Mitwirkung von PatientInnen an der Entwicklung einer dermatologischen Therapie-App – ein Bericht aus der Praxis
Anne Koopmann, Anna Maria Pfeifer, Lara Schweickart, Nathalie Biniaminov, Valentin Haas, Philipp Marquardt, Astrid Gößwein, Christopher Czaban, Sergey Biniaminov, Mara Blauth, Caroline Glatzel, Christoph Zimmermann, Wilhelm Stork, Victor Olsavszky, Astrid Schmieder • Die Dermatologie • 2024
Benefits of participatory involvement of patients in the development of a dermatological treatment app—A report from the practice
Complement Convertases in Glomerulonephritis: An Explainable Artificial Intelligence-Assisted Renal Biopsy Study
Wiech, Thorsten; de las Mercedes Noriega, Maria; Schmidt, Tilman; Wulf, Sonia; Koch, Timo; Marquardt, Philipp; Biniaminov, Sergey; Hoxha, Elion; Tomas, Nicola M.; Huber, Tobias B.; Zipfel, Peter F. • Journal of the American Society of Nephrology • 2021
An AI assisted study on complement convertases in glomerulonephritis using renal biopsies.
Projects
Deep Learning Framework for Training Classification, Segmentation, Object Detection and Instance Segmentation Models
Deep Learning Framework for Training Classification, Segmentation, Object Detection and Instance Segmentation Models
Microscopy Image Viewer
A full-stack application for viewing and analyzing microscopy images.
C# Multiplayer Framework for VR/AR Applications in Unity
A multiplayer framework written in c# for vr and ar applications. Can be used in Unity to enable TCP and UDP based multiplayer sessions with various features.
MMWrapper
A wrapper around the ecosytem of OpenMMLab to easily train models using an easy to use config
Mulitmodal Emotion Detection
A project that uses the RAVDESS Audio and Video dataset to train multimodal models that can detect emotion
Time Logging Project
Time Logging Application to Manage Employees Working Times and Create Monthly Reports. Written in C# WPF
A simple Machine Learning Trainer and Annotation GUI in C#/WPF
A simple C#/WPF application do define a custom deep learning architecture. Was extended to include a simple annotation tool for standard image formats.
Document Parser using Language Models
Parse a Document in a structured way using LLMs
Computer Science Master at the Karlsruhe Institute of Technology
Semester 1
Deep Learning for Computer Vision II: Advanced Topics
Learn about newest reserach topics in computer vision
Natural Language Processing
Learn everything about NLP from the ground up starting from basic morphology and ending with current state of the art llms
Energy Informatics 1
Learned about energy forms, storage, transmission, and conversion; the use and evaluation of equations; energy system components; energy informatics applications; analysis of the German energy system; energy economics; and the Smart Grid concept. The module covered energy forms, systems, storage, power plant processes, renewable energies, energy transmission networks, future electrical networks, and energy economics.
Semester 2
Machine Translation
Learned about linguistic approaches to machine translation, with a focus on methods and algorithms for statistical machine translation (SMT), including word alignment, phrase extraction, language modeling, decoding, and optimization. Explored methods for evaluating machine translations and examined applications of machine translation through the example of simultaneous speech-to-speech translation. Practically applied the acquired knowledge by training a translation system during exercises.
Practical Course Computer Vision for Human-Computer Interaction
Created a graph-explaining system for visually impaired people.
Machine Learning for Natural Sciences Exercises
Semester 3
Humanoid Robots - Seminar
Wrote a paper about newest trends in Movement Primitives
IT Security
Learned advanced topics in cryptography and IT security, including sophisticated techniques and security primitives to achieve protection goals. Gained an understanding of scientific evaluation and analysis methods for IT security, such as game-based formalization of confidentiality and integrity, and concepts of security and anonymity. Acquired knowledge about data types, personal references, legal, and technical foundations of data protection. Learned the basics of system security, including buffer overflow and return-oriented programming. Explored various mechanisms for anonymous communication (TOR, Nym, ANON) and evaluated their effectiveness. Understood blockchains and their consensus mechanisms, assessing their strengths and weaknesses.
Machine Learning in Climate and Environmental Sciences
This module covers key concepts for real-world applications of machine learning, focusing on environmental data science. Topics include the foundations of machine learning (such as the curse of dimensionality, cross-validation, cost functions, and feature engineering), widely applied regression, classification, and unsupervised learning algorithms (like LASSO, random forests, Gaussian processes, neural networks, LSTMs, transformers, and self-organizing maps), time series forecasting, and causal inference. It also explores explainable AI methods (such as SHAP value analyses, feature permutation methods, and intrinsically interpretable methods).
Semester 4
Advanced Machine Learning and Data Science
Develop a system to regress measures of option pricing before and after ecb meetings to determine their impact on the price
Master Thesis
Few Shot Image to Image Translation