Professionelle Machine Learning Neuronale Netze Experte Data Scientist AI KI Developer Mathematiker Physiker Berlin
- Verfügbarkeit einsehen
- 0 Referenzen
- 95€/Stunde
- 13055 Lichtenberg
- DACH-Region
- de | en
- 15.03.2021
Kurzvorstellung
Qualifikationen
Projekt‐ & Berufserfahrung
5/2020 – 8/2020
Tätigkeitsbeschreibung
Developing an OCR for Invoice recognition software which checks for errors in the Invoice
• CombiningclassicalImageRecognitiontechniqueswithmodern Neural Networks approaches
• DataMining,DataTracking
• UtilizingmathematicalStatisticsforDataPreparationandData Cleansing
• WritinganefficientSQL-Commandsearchfunctiontomakethe work easier
• UtilizingImagerecognition
• UtilizingDeepLearningTechniquesforTextRecognitiononthe invoice
• BuildingDeepAuto-encodersforthedimensionalityreduction and denoising of the Data
• Developing the Neural Network for the Supervised Learning with custom losses and custom layers
• UsingReinforcementLearningTechniqueswhichsuggeststhe worker what to do in the next step
• Utilizing Convolutional Neural Networks (CNNs) for better Logo detection
• Testingthesoftwareintherealusecase
• Transportingthecodetothewebserver,settingupanefficient SSH Connection and local Cloud Computing infrastructure
Statistiken, Maschinelles Lernen, Torch, Text Mining, Text-Extraction, Textklassifikation, Python
3/2020 – 5/2020
Tätigkeitsbeschreibung
Project details:
Developing AI for a medicine application (Parkinson disease) which suggests the optimal behavior for the patient to reach and boost
their healthcare goals.
Coupling Supervised Learning with Reinforcement Learning to enable online learning of patients needs and make instant
suggestions to the user
-Data Mining, Data Tracking
-Utilizing mathematical Statistics for Data Preparation and Data Cleansing
-Creation of synthetic data to boost the AI and find the principal architecture of the AI which learns our original data more
precise
-Check with the synthetic data what is needed to couple the Supervised Learning AI and the Reinforcement Learning AI
-Building Deep Autoencoders for the dimensionality reduction and denoising of the Data
-Developing the Neural Network for the Supervised Learning with custom losses and custom layers utilizing self-invented and more efficient custom Neural Networks architectures in Tensorflow
-Developing the Neural Network for the Reinforcement Learning with custom losses and custom layers
-Utilizing Recurrent Networks Architecture (RNNs) for time series combined with Convolutional Neural Networks (CNNs)
-Connecting the Supervised Learning and the Reinforcement Learning Networks to each other to simulate the patient and doctor relations, which helps the AI to make realistic and useful suggestions to the Patient
-Transporting the AI to the Webserver, setting up a local Cloud Computing infrastructure with advanced GPUs and efficient SSH
Connection
-Building BigData Pipelines with Hadoop and Spark to allow parallel and cluster computing which helps to find the better
appropriate Neural Network models
Qualifications:
Neural Networks (CNNs, Recurrent Nets, RNNs),
Machine Learning,
Supervised Learning,
Reinforcement Learning,
Statistics,
Python,
R,
Tensorflow,
Numpy
BigData,
Spark,
Scala,
Hadoop,
PySpark,
Kafka,
Git,
Bitbucket,
Docker,
Bash,
AWS,
Azure,
Cloud Computing,
DataClustering,
Databricks,
Tinc VPN
Jira,
Confluence,
Trello,
Jupyter Lab,
Jupyter Notebook,
Spider,
Visual Code
Statistiken, Apache Hadoop, Big Data, Faltendes Neuronales Netzwerk (CNN), Rekurrentes Neuronales Netzwerk (RNN), Tensorflow, Apache Spark, Microsoft Azure, Amazon Web Services (AWS)
1/2020 – 4/2020
Tätigkeitsbeschreibung
Project details:
Consulting for an SAP Addon which is predicting the right quota of the bills making the contingation process easier for the employee,
the Addon customizedly learns the way of working of the employee and makes more and more better predictions which bills belong
to which contingent.
Developing one simple and one more complicated prototypes
Coupling Reinforcement Learning and Supervised Learning techniques to each other making past data and current data more
efficient for the on fly learning
Data Preparation, Data Clustering using EM, ICA, PCA techniques
Utilizing mathematical Statistics for unsupervised learning and Clustering methods
Dimensionality Reduction of the Data with Deep Autoencoders, using denoising techniques in the Autoencoder by
programming own custom Tensorflow Blocks
Writing Python and JavaScript APIs to enable the transfer between the SAP Addon and the Python Script with the AI
Dockerizing the Anaconda environment, setting up a virtual environment in the server
Utilizing BigData Technologies like Spark to build fast and efficient cluster computing environment
Utilizing Kafka Stream APIs to enable fast and efficient learning between the Reinforcement and the Supervised Learning AIs
form the one side and the data transfer from user to AI from the other side
Utilizing Cloud Computing platforms like AWS to boost the training procedure of AIs
Qualifications:
Mathematical Statistics,
Artificial Intelligence,
Neural Networks,
GANs,
RNNs,
CNNs,
Machine Learning,
Supervised Learning,
Unsupervised Learning,
Reinforcement Learning,
Statistics,
Python,
R,
Tensorflow,
Numpy
BigData,
Spark,
Kafka,
Git,
Bitbucket,
Docker,
Bash,
AWS,
Cloud Computing,
DataClustering,
Databricks,
Trello
Auswahl Vertretern / Kooperationspartnern, Statistiken, Apache Hadoop, Big Data, Faltendes Neuronales Netzwerk (CNN), Maschinelles Lernen, Neuronale Netze, Pytorch, Tensorflow, Überwachtes Lernen, Docker, Python, Microsoft Azure, Amazon Web Services (AWS)
12/2019 – 1/2020
Tätigkeitsbeschreibung
Project details:
-Used top modern non-relational webdeveloping tools VueJS and NodeJS
-Using HTML, CSS and JavaScript
-Using non-relational Databases MongoDB
-Using NodeJS for serverside programming and for data processing of users, writing Email and CV APIs with the help of NodeJS
-Configuring a Strato Virtual Server as a Ubuntu Server to be able to use the efficient Linus environment
Mongodb, Ubuntu, CSS (Cascading Style Sheet), HTML, JavaScript
9/2019 – 12/2019
Tätigkeitsbeschreibung
Project details:
-Adapting psychological Models to the AI, AI learns to make customized suggestion to the consumer based on its behavior.
-Creating synthetic data to build real and fake personalities
-Using GANs to generate more synthetic data, making unrealistic data that the Neural Net later can distinguish between the original and the fake one making the Nets more robust on adversarial examples
-Data Mining, Data Tracking, Data analysis/preparation/optimization
-Using Deep Autoencoders to denoise the Data
-Data clustering/ dimensionality reduction with statistical tools, self written code+python packages Pandas, using R
-Design and creation of appropriated Neural Nets, Hyperparameter tuning
-Heavily used my mathematical thinking and knowledge to choose the right optimization techniques, using very innovative
ideas on Neural Networks which I developed in may master's thesis
-Usage of modern automated Tools like AutoML to run the Hyperparameter tuning more autonomous, using grid techniques for the searching processes also for the activation functions
-Usage of cloud computing platforms: AWS, Azure
-Write, Run and Debug self written custom Neural Network codes, using different ML Libraries: TensorFlow, PyTorch
-Successful validation of self written Neural Nets, new result in the theoretical psychology
Qualifications:
Data Science,
Big Data,
Predictive Analytics,
Machine Learning,
Statistics,
Neural Nets,
Python,
Pandas,
CUDA,
Tensorflow,
PyTorch,
Numpy,
Spark,
Hadoop,
Kafka,
Go,
C++,
Java,
SQL,
AWS,
Azure
Data Science, Statistiken, Apache Hadoop, Big Data, Microsoft SQL Server Analysis Services (SSAS), Predictive Analytics, Java Database Connectivity, Maschinelles Lernen, Neuronale Netze, Pandas, Tensorflow, Apache Spark, Python, Microsoft Azure, Amazon Web Services (AWS)
5/2019 – 10/2019
Tätigkeitsbeschreibung
Project details:
-Writing Python APIs, writing Data Pipeline scripts, configuring servers, installing Docker containers and Anaconda environments in the server to enable Data Science APPs and Projects to run on it more efficiently
-Helping with the data mining procedure with the team of the justplan aktiv
-Utilizing my advanced mathematical knowledge in BigData analysis, preparation and optimization of the training data for Neural Networks
-Data denoising/ clustering/ dimensionality reduction with statistical tools, self written code+python packages Pandas, using R
-Gaining first good results with my self developed innovative Neural Networks, solving a 2 years lasted problem with unifying of several Neural Networks in one single Network
-Design and creation of appropriated Neural Nets, Hyperparameter tuning
-Usage of modern automated Tools: AutoML
-Usage of cloud computing platforms: AWS, Azure
-Write, Run and Debug self written custom Neural Network codes, using different ML Libraries: TensorFlow, PyTorch
-My team gained the better results on the same data than the other team, the code is also faster than the one of the other
team
Qualifications:
Data Science
Big Data
Predictive Analytics
Machine Learning
Neural Nets
Python
Pandas
CUDA
Tensorflow
PyTorch
Numpy
Azure
AWS
Java
Go
JavaScript
Spark
Hadoop
Kafka
C++
SQL
Data Science, Apache Hadoop, Big Data, Microsoft SQL Server Analysis Services (SSAS), Predictive Analytics, Java Database Connectivity, Neuronale Netze, Pandas, Pytorch, Tensorflow, Apache Spark, C++, Python, Microsoft Azure, Amazon Web Services (AWS)
10/2018 – 5/2019
Tätigkeitsbeschreibung
-Working as a lecturer in areas Mathematics, Physics and Machine Learning.
-Giving Python and Machine Learning workshops for firm employees
-Giving private lessons in Mathematics,StatisticsandProgramming, giving special -Python programming for Artificial Intelligence for beginners and intermediates
-Being a guest lecturer at the National Polytechnic University of Armenia in Machine Learning and Statistics
Mathematik, Physik, Statistiken, Microsoft SQL Server Analysis Services (SSAS), Java Database Connectivity, Maschinelles Lernen, Neuronale Netze, Pytorch, Tensorflow, C++, Python
2/2017 – 7/2018
Tätigkeitsbeschreibung
Project details:
-Helping the Team of the AI department to extend the Machine Learning cluster, to collect and prepare the vast amount of data, consulting about the utilization of the information in the collected data, building various small APIs to enable the working between the teams of the different IT- departments of Siemens
-Utilize my mathematics knowledge to build Statistical Python Scripts for Data Preparation
-Building Hadoop pipelines for the large available data of Siemens
-Utilizing various statistical processing methods like Autoen- coders, EM, tNC, PCA, ICA algorithms
-Doing Scientific research,reading,understanding and showing relevant Papers to the team
Mathematik, Statistiken, Java Database Connectivity, Maschinelles Lernen, Neuronale Netze, Pandas, Pytorch, Tensorflow, Apache Spark, C++, Python
Ausbildung
Berlin
Berlin
Über mich
Hallo,
ich bin Mathematiker aus Berlin und ein begeisterter Neuronale Netze Entwickler. Ich habe mich in den letzten Jahren ausgiebig in dieses Gebiet vertieft sowohl theoretisch im Rahmen meines Masterstudiums als auch programmtechnisch bei diversen Freelance Projekten. In meinem Masterstudium habe ich mich über eine sehr innovative Methode beschäftigt wie man Neuronale Netze als Differentialgleichungen ansieht was uns dann ermöglicht die gesamten mathematischen Tools von den letzten 200 Jahren auf Neuronalen Netzen anzuwenden. Dies ermöglicht viel interpretierbare und robustere Netze zu bauen, was momentan z.B. ein großes Thema bei selbstfahrenden Fahrzeugen ist. Mein breites Wissen in der Anwendungsmathematik ermöglicht es mir jedes Machine Learning/Big Data Problem professionell zu lösen und ggf. problemspezifisch fortgeschrittenere Algorithmen zu entwickeln. Um diesen Level zu erreichen habe ich fast das gesamte Feld der Anwendungsmathematik studiert: Numerics, Finite Element Methods, Optimisation Theory, Differential Equations, Statistics/Stochastics. Dieses Wissen hebt mich von den anderen Data Scientists ab, welche zwar Programmmodule bedienen können, allerdings kein tiefes Wissen haben um ein ernsthaftes Problem zu lösen, was zu oft auftritt und man begnügt sich mit suboptimalen Lösungen. Mein Ansatz ist hier etwas anderer: wenn nötig, greife in die Programmmodule ein und modifiziere dort den Programmcode wie z.B. Implementierung von custom layern in Tensorflow oder PyTorch, beides beliebte Neuronale Netze Module. Weiterer Ansatzpunkt von mir ist, nicht mit Kanonen auf Spatzen zu schiessen, sondern für jedes Problem die maximal einfache und effiziente Lösung zu finden, wofür ich das nötige Wissen habe. Das spart Zeit, Geld und Nerven. Um auf top aktuellem Niveau zu bleiben, besuche ich ständig Workshops und diskutiere aktiv mit meinen Kollegen. Ich bedanke mich bei Ihnen über mich gelesen zu haben und ich beantworte gerne jede Frage die an mich gestellt wird.
For English speakers:
Hello,
I am a mathematician from Berlin, Germany, and an excited Neural Networks programmer. I wrote my maths master's thesis on Neural Networks and Differential equations showing that Neural Networks are Differential Equations that help us to use the whole maths theory to find better and more stable Neural Networks being robust against adversarial examples. I love to go inside of the Machine Learning programs and understand every part of it, why it works well or bad and with the help of my immense mathematical understanding to improve it, if necessary. For this, I studied the whole bunch of applied Mathematics: Finite Element Methods, Optimisation Theory, Differential Equations, Statistics and Stochastics. This knowledge picks me up on others who only use the standard machine learning packages and are not able to create really robust and stable networks. Besides of that I love coding and implementing my ideas on Neural Networks if necessary improve Google's Tensorflow or PyTorch by writing APIs if they are not advanced enough. I constantly improve my Machine Learning/Big Data knowledge by active programming and visiting various workshops, seminars and university courses where I am actively discussing methods with my colleagues. That's about me, I will be thankful for every offer and for every question I will be glad to answer it!
Weitere Kenntnisse
• Python, Go (GoLang), Java, C++
• NumPy,Tensorflow, Keras, PyTorch
• Convolutional Neural Networks (CNNs)
• Recurrent Neural Networks (RNNs)
• Long Short-Term Memory Networks (LSTM)
• Scikit-learn
BigData
• Scala
• Databricks, Spark, PySpark, Hadoop, Kafka
• Python Pandas
• SQL, NoSQL, MySQL, MongoDB
• Docker, Kubernetes, Git, Jira, Confluence, Bash
Matlab
• Statistics and Machine Learning Toolbox
• Optimization Toolbox
• Curve Fitting Toolbox
• Partial Differential Equation Toolbox
Cloud Computing
• AWS
• Azure
• Google Cloud
• Cuda, Cudnn
Frontend Development
• HTML & CSS
• JavaScript
• VueJS
• ReactJS
• Angular
Backend Development
• NodeJS
• RESTful API
Other skills:
Latex
• Beamer
• PGF/TikZ
MS Office
• Word
• Excel
• Powerpoint
Vim, Nano
Linux, Mac OS, Windows
Trello, Jira, Bitbucket
Programming Way: more than 10 years of Coding Experience
2009: Choosing Compute Science school classes, building my first applications with the programming language "Delphi"
2010: Programming small HTML Website for the School
2011: Choosing CAS Mathematics, solving mathematical algorithms on the "Derive" Computer Algebra System, writing own solvers for square root problems
2012:WritingownregressionApptofitthemeasurements at my physics experimental lectures
2013: Starting to learn Python for the University courses, writing practical Apps with my university college to make easy to solve our home tasks
2014: Developing a website with HTML, CSS and JavaScript for my Professor
2015: Writing own Finite-Element realistic Simulation software in Matlab, which requires a very strong mathematical intuition and knowledge
2016: Writing FEM Simulation Codes in Python, utilizing very modern Matrix-Vector Linear Algebra Computational C++ Modules
2017: Utilizing Jupyter Notebook, Latex and Python to write my Bachelor’s thesis in Mathematics, which was 70% coding, found numerical bad behavior in the theory of the mathematics paper used for my Bachelor’ thesis, for which I was rewarded from my professor
2017:AccomplishingMachineLearningIattheTechnische Universität Berlin
2018: Accomplishing Machine Learning II at the Technische Universität Berlin
2018: Accomplishing Deep Learning Courses at the Freie Universität Berlin
2018: Accomplishing Python Programming for Deep Learing Courses at the Technische Universität Berlin
2019:AccomplishingDeepLearningattheTechnischeUniversität Berlin
2019: 11 months permanent writing Python, Matlab, Java Codes for my Master’s Thesis in Mathematics, more than 50% of the work was the programming part, gained with my numerical experiments new results in the area of Neural Networks and Differential Equations
Persönliche Daten
- Deutsch (Muttersprache)
- Englisch (Fließend)
- Europäische Union
- Schweiz
Kontaktdaten
Nur registrierte PREMIUM-Mitglieder von freelance.de können Kontaktdaten einsehen.
Jetzt Mitglied werden