# Dr. Mario Michael Krell > Leader | Senior Machine Learning Researcher | Postdoc | Teacher | Mathematician | Problem Solver | Scientific Paper Writer | Mentor | Data Scientist | DFKI, UC Berkeley, Mercedes, Graphcore, Optum, Annapurna Labs Location: San Francisco Bay Area, United States Profile: https://flows.cv/drmariomichaelkrell I combine analytical strengths and mathematical skills with cross-functional teamwork and leadership skills to achieve business goals in machine learning research. I am an innovative leader, leveraging cumulative experience across diverse applications gained in individual contributor and management roles to deliver best-fit solutions to complex problems. I draw on communication strengths to collaborate in multicultural cross-functional teams (engineers, computer / neuroscientists, managers, POs). My core competencies include classical machine learning, deep learning, evaluation, framework development, testing, documentation, optimization, project management, and presentation. I have strong problem solving, mathematical analysis, interdisciplinary teamwork, and leadership skills. I am a senior Machine Learning enthusiast with 15 years of cumulative experience in a variety of applications like AI accelerators, automotive, multimedia, brain-computer interfaces, and robotics. See my google scholar profile: https://scholar.google.de/citations?user=77LZ0ckAAAAJ. I would like to lead machine learning research and development for challenging applications that help humanity. SKILLS - problem solving, machine learning, deep learning, mathematical analysis, optimization - leadership experience with up to 8 direct reports, stakeholder interaction, and as scrum master/PO - collaboration in multicultural/interdisciplinary teams (engineers, computer/neuro-scientists, manager, PO) - independent research and data analysis (>40 publications, >500 citations), teaching, scientific presentation - software development: Python, TensorFlow, PyTorch, NumPy, Git, Sphinx, HPC, documentation, PySpark - Machine Learning Models: GNN, SVM, SVR, CNN (ResNet-50), NLP (BERT), probabilistic models (ABC, AdGMoM), RL (MiniGo), evaluation, capacity, differential privacy, signal processing, clustering - basic application knowledge: brain-machine interfaces, robotics, multimedia, automotive, hardware acceleration - Multilingual: English (fluent), German (fluent), French (basic) ## Work Experience ### Sr. Software Engineer - AI/ML @ Amazon Jan 2024 – Present | Cupertino, California, United States Lead the efforts building distributed training and inference support into Pytorch, Tensorflow, Jax using XLA and the Neuron compiler and runtime stacks. This role will help tune these models to ensure highest performance and maximize the efficiency of them running on the customer AWS Trainium and Inferentia silicon and the TRn1 , Inf1 servers. ### Principal Machine Learning Engineer @ Optum Jan 2023 – Jan 2024 | San Francisco Bay Area - Improved modelling performance from close to 0 to around 96% hit rate according to business to optimize their processes. - Designed and implemented modelling framework from ground up and brought it to production. - Implemented modular framework for feature calculation and aggregation to consider different time horizons, statistics, and selection criteria. Framework could be understand by non-experts. Boosted number of features from 38 to more than 6500. - Implemented modelling pipeline to predict and measure drug efficiency. - Implemented modular framework for table transformation ML pipelines to enable version tracking, logging, and application for inference. Tuned performance of pipeline. (PySpark, Databricks, MLFlow) ### AI Engineering Manager | Machine Learning Research Manager @ Graphcore Jan 2022 – Jan 2022 | Palo Alto, California, United States I managed up to 8 direct reports, including change management, teaching, and coaching. I handled human resources / people management issues, such as salary assessment, leave, and performance review. - Drove product delivery, communicating 1-on-1 with stakeholders, connecting right people to each project, strategizing on plan and execution, and deescalating and ramping down projects to achieve business and customer objectives. - Defined a new department structure, explaining and implementing changes gradually to gain buy-in and increased organizational efficiency to reduce time from customer engagement to product delivery. ### Principal Machine Learning Lead | Artificial Intelligence Applications Specialist @ Graphcore Jan 2019 – Jan 2022 | San Francisco Bay Area I collaborated with customers and colleagues to accelerate customer and benchmark models based on implementation, analysis, and optimization skills. I documented research and approaches for marketing. - Boosted sales activity by delivering improved performance in alignment with main competitor, including marketing material used in almost every customer interaction: deep learning performance benchmark (MLPerf) submission with 3.2x speed-up in ResNet-50 and 2x speed-up with packing in Natural Language Processing. - Tuned 8 customer models on IPU (GNNs, image processing, prob. models, RL [MiniGo]), using software development, problem solving, and communication skills, which enabled public applications for marketing and sales team interaction. - Wrote and presented 11 scientific papers, documentation, performance guides, scaling guide, and blogs, drawing on research and writing skills to create marketing and educational material that promoted company hardware as tool for accelerating research and educating customers and colleagues on how to optimize applications. - Organized 1-2 team meetings per week to educate and encourage collaboration for better motivation and deliverables. - Recognized as top performer for supporting business mission achievement, including driving profitable sales growth, serving as role model to others, and building internal and external partnerships across functions and at all levels, 2021. ### Principal Machine Learning Scientist | Principal Data Scientist @ Mercedes-Benz Research & Development North America, Inc. Jan 2018 – Jan 2019 | Sunnyvale I managed up to 3 direct reports and 4 big data analytics / predictive algorithms projects to support driver display interaction. - Wrote 2 white papers and 10 patent proposals (5 filings), documenting current and future innovation to protect knowledge and ideas and future revenue. - Analyzed large-scale customer data, verifying data for sales use and creating POC for data usability, including existing part specification verification for quality team. - Improved the customer car experience by fixing mathematical errors with my analytical skills and implementing evaluation frameworks and new predictive algorithms. ### Postdoctoral Research Scholar @ UC Berkeley Jan 2017 – Jan 2017 | International Computer Science Institute I guided 5 undergraduate students for the undergraduate research apprentice program. I used multimedia commons to implement big data study tools across disciplines. I explored deep learning applications. - Conducted / published deep neural network empirical capacity analysis, increasing generalization capability understanding. - Implemented a deep neural network for EEG data classification, supervising follow-up with student master thesis that showed improved predictive algorithm applicability in transfer learning across subjects. ### Postdoctoral Researcher in Machine Learning @ University of Bremen Jan 2015 – Jan 2017 I consulted in all machine learning projects. I acquired and worked in 4+ projects. - Educated on machine learning to enhance project results, including facilitating machine learning lectures, supervising students, and organizing machine learning and signal processing workgroups. - Worked in different projects like RECUPERA-Reha, BesMan, Entern, Robocademy - Supported projects with or for the industry (health devices, cars, airplanes, wearables, telecommunication, art, employer’s liability insurance association) - Helped with project acquisition (contribution to more than 10 project proposals). - Improved algorithms to better operate when few data or few resources are available. - Led ML software development: pySPACE (open-source release, refactoring, documentation, user support, user interface, multi-class, regression, pipeline decoding visualization), allowing colleagues to run experiments efficiently, distributed, and independently. - Supervised master thesis (data selection strategies for SVMs; automatic processing chain optimization with pySPACE). ### Machine Learning Researcher @ University of Bremen Jan 2010 – Jan 2015 | Bremen • Developed general concepts for connecting SVM variants (regression, one-class classification, online learning), enabling 4-in-1 lecture introduction without overwhelming students by improving teaching and usability understanding. • Developed backtransformation, a new visualization concept for signal processing chains. • Developed new parameter optimization algorithms and sensor selection algorithms to improve the classification outcome. • Implemented the open source release of pySPACE (improving, refactoring, and restructuring) • Improved documentation by developing a documentation concept and using an automatic API documentation generator to simplify changes by future developers. • Developed a generic unit testing framework in pySPACE to easier spot errors. • Implemented numerous further improvements to pySPACE (e.g., multi-class classification, regression, bug fixes, style improvements, data handling). • Supervised student assistants and a master thesis ### Junior Machine Learning Researcher @ DFKI GmbH (German Research Center for Artificial Intelligence), Robotics Innovation Center Jan 2009 – Jan 2010 | Bremen, Germany Implemented, tested, and evaluated machine learning algorithms on electroencephalographic data. Wrote project proposals. • Developed a mathematical model for space simulation with a six-axis industrial robot and a cable-guided 3D-movement system to enable the representation of realistic docking maneuvers. ### Student Assistant @ Humboldt-Universität zu Berlin Jan 2005 – Jan 2009 | Berlin - Corrected exercise sheets to ensure, content was learned correctly and proof skills are improving. ## Education ### Doctor of Philosophy (PhD) in Computer Science (Machine Learning) University of Bremen ### Diploma in Mathematik Humboldt-Universität zu Berlin ### Pre-Diploma in Computer Science Humboldt-Universität zu Berlin ### university-entrance diploma in Mathematics and Computer Science, General Education Carl-Friedrich-Gauß-Gymnasium ## Contact & Social - LinkedIn: https://linkedin.com/in/drmario - Website: http://mmkrell.github.io/CV/index.html - Website: https://medium.com/@dr.mario --- Source: https://flows.cv/drmariomichaelkrell JSON Resume: https://flows.cv/drmariomichaelkrell/resume.json Last updated: 2026-04-01