About Maximilian
Ich entwickle und implementiere produktionsreife KI-Systeme – von der Entwicklung bis zum (Edge-)Deployment.
Schwerpunkte:
- Real-time Object Detection & Tracking
- Andere Computer Vision Aufgaben (z.B. Segmentation, Depth Estimation, OCR und Image Retrieval, etc.)
- Voice Agents, LLM adaptation and integration, Document Retrieval
- Modell-Optimierung & Quantisierung für schnelle Laufzeiten (NVIDIA Jetson, TensorRT, ONNX)
- MLOps & skalierbare Trainings-Pipelines (Kubernetes, MLflow, multi-GPU)
- Implementierung von AI-Web-Anwendungen und APIs (z.B. auch Voice Agents, Tooling Plattformen, u.v.m.)
Mit über 8 Jahren Erfahrung in der Entwicklung sicherheitskritischer KI-Systeme und der Leitung des kompletten 2D-Object-Detection- & Tracking-Stacks in meiner Abteilung bei Hensoldt entwickle ich robuste, produktionsreife Computer-Vision-Lösungen.
Ich integriere modernste KI-Module (Object Detection, Segmentation, Depth Estimation, OCR, Motion Detection, LLMs, Voice Systems u. v. m.) nahtlos in alle Arten von Systemen – von Edge-Geräten über Cloud-Infrastrukturen bis hin zu Web-Applikationen, Mobile Apps und bestehender Unternehmenssoftware. Über REST-APIs, skalierbare Backend-Architekturen und Edge-to-Cloud-Pipelines.
Ergebnis: Performante, wartbare und sofort einsetzbare KI-Systeme – schnell, pragmatisch und auf höchstem technischem Niveau.
Aktuell nebenberuflich verfügbar (ca. 8-10 Std/Woche) für:
Beratung & Architektur-Reviews
Modell-Training & Optimierung
Prototypen bis Produktion
Implementation einer AI-(Web)App
Sprachen: Deutsch (Muttersprache), Englisch (verhandlungssicher).
German
Native or bilingual
English
Fluent
Remote only
Primarily works remotely
Experience
- HensoldtFull-time professionalOctober 2021 - Today (4 years and 9 months)Munich, Germany
- Take lead and ownership for the development of the complete 2D object detection and tracking software stack, from research to deployment
- Manage and curate internal training datasets, including defining annotation strategies and overseeing labeling workflows
- Implement cutting-edge real-time object detection architectures and adapt them to production scenarios
- Design and implement customized, problem-specific deep learning algorithms for various computer vision tasks on edge devices like NVIDIA Jetson using C++ and C
- Key role in designing, deploying, and maintaining a multi-node, GPU-accelerated Ku bernetes cluster, ensuring it was fully optimized and user-friendly for machine-learning pipelines
- Develop a large-scale deep learning training pipeline using modern MLOps practices (Kubernetes, multi-server training, MLFlow) to train competitive, task-specific computer vision models on large datasets
- Built and deployed motion-based algorithms for the detection of tiny moving objects, specializing in drone identification
- Quantize deep learning models to optimize inference speed for edge deployment (not object-specific) to accelerate the labeling process
- Design and implement an image retrieval system based on feature-space similarity
- Train and deploy high-performance panoptic segmentation models with strict latency constraints
- Train and deploy depth estimation models optimized for real-time inference
- Implement an OCR system supporting multiple languages efficiently
- Design and operate scalable model-serving infrastructure using Kubernetes Pods and Triton Inference Server, delivering multiple ML models via REST APIs to customer-facing OSINT applications
- Regularly review state-of-the-art research to inform architectural and strategic decisions
- Develop a semi-supervised deep learning model for radar-based object identification
- Define tasks for, interview and supervise Master students, ensuring alignment with team objectives and research quality
- Fraunhofer IKSFull-time professionalMarch 2019 - September 2021 (2 years and 6 months)Munich, Germany
- Identifying, extenting and integrating promising methods for improving safety, explainabil ity and reliability of autonomous systems which are utilizing Machine Learning methods
- Reproduction and evaluation of state-of-the-art methods for quantifying uncertainty and detecting out-of-distribution examples in Deep Neural Networks applied on Computer Vision problems (Classification, Segmentation, Keypoint Estimation, 2D/3D Object Detection)
- Working in compliance with functional safety standards such as SOTIF(ISO/PAS 21448) Development of new methods for uncertainty quantification in 2D and 3D Object Detection
- Integrating developed methods into demonstrators and proof-of-concept prototypes
- Publication and presentation of scientific papers
- Supporting business development on customer acquisition from the industry
- Communicating research with different departments including marketing
- Participation in interviewing potential new employees for the department
- Supervising working students and utilizing their contributions
- Participation in events and meetings for exchanging ideas with different research institutes and industry
- Leading a research project end-to-end, providing strategic direction, coordinating tasks of team members, and ensuring delivery of high-quality, technically sound outcomes
- Developed and authored competitive research proposals and funding applications, trans lating complex AI concepts into clear, compelling narratives aligned with sponsor priorities
- Managed project operations and administrative processes including budgeting, reporting, compliance, and stakeholder communication to ensure smooth execution and sustained project momentum
Recommendations
Be the first to recommend Maximilian
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- BSc Engineering ScienceTechnical University of Munich2017BSc Engineering Science
- MSc Artificial IntelligenceThe University of Edinburgh2018MSc Artificial Intelligence