Andrzej Gajda, Ph.D. Email & Phone Number

Utrecht, Ut, Nl

Utrecht, Netherlands

At the beginning I was helping with development and maintenance of Geval and Gonito projects (both written in Haskell), which together serve as an open source platform for machine learning models competitions.Currently I am building a backend in Python that is based on the two previous projects, where the new frontend is written in React. Application is contenerized with PostgreSQL serving as a database.

Arnhem-Nijmegen Region

At TenneT I help with the analysis of large amounts of unstructured data (scanned pdf documents), where we try to extract valuable information, and at the same time make the content of this unstructured data available to other users as a search engine solution. This involves maintenance of a kubernetes cluster in Azure with Elasticsearch as the search engine, and tools that involve AI algorithms to extract information of value from documents stored in Azure blob storage.In addition, I also help refactor and maintain an already existing solution for intelligent document checking, that is used in TenneT to check documentation for large projects.My main tasks involve:- Creating tools in Python for content extraction and classification from scanned documents, and storing information in Elasticsearch.Technologies: Tesseract OCR, Elasticsearch, Kubernetes, Python, Azure cloud- Using and fine-tuning algorithms for content extraction and documents classification, e.g. Donut model.Technologies: Python, Transformers, Azure Machine Learning Studio- Code refinement and maintenance in a large project for documentation checking.Technologies: Python, Poetry, Kubernetes, Argo workflows

Zeist, Utrecht, Netherlands

As an AI engineer at Y I am mainly responsible for developing a chatbot using Rasa framework and graph knowledge base along with LLMs. I also participate in a project of intelligent search engine. Both solutions are developed in Azure cloud and hosted in a kubernetes cluster.My main tasks were:- Code development and refinement along with deployment of parts of the systemTechnologies: Python, Skaffold, Ansible, Kubernetes, Docker, Kind- Using GPT api and deployment of Llama2 on virtual machines in Azure and in Amazon cloudTechnologies: Azure / Amazon VM with GPU usage for inference, Python, prompt engineering- Building images for Docker containers with GuixTechnologies: Guix, Docker, Python

Poznań, Wielkopolskie, Poland

As a Senior Data Analyst in GSK I've been working with machine learning models to improve classification of internal incidents in the company and moving data from ServiceNow to Azure.- Multiclass classification: determination of 5 parameters for written request based on the description of the problem (Global Employee Recognition award)- Automation of finetuning models in given time intervals and preparation of metrics calculations for their performance measurement- Developing Virtual Agent utility for Natural Language Understanding and Processing (ServiceNow, IBM Watson)- Moving data from ServiceNow to Azure SQL server

Classes I taught:- Artificial Intelligence and Machine Learning- Reasoning- Introduction to Haskell Programming- Introduction to Python Programming- Mathematical Foundations of Cognitive ScienceI have promoted 5 BA theses and 1 MA thesis. The scope of the work involved mainly: Artificiall Neural Networks, programming in Python and Haskell, Logic Programming.

Poznań, Wielkopolskie, Poland

The project concerned Modelling of Abductive Reasoning and was funded by the National Science Centre in Poland (Sonata-Bis grant, DEC-2013/10/E/HS1/00172). My role in the project was to develop a script written in Haskell and Python that allowed us to model abductive reasoning in a neural-symbolic system (i.e. a system that combines logic programs and artificial neural networks).

Poznań, Wielkopolskie, Poland

The project concerned Modelling of Abductive Reasoning and was funded by the National Science Centre in Poland (Sonata-Bis grant, DEC-2013/10/E/HS1/00172). My role in the project was to develop a model of abductive reasoning in a neural-symbolic system (i.e. a system that combines logic programs and artificial neural networks). Part of the work involved proving that using Immediate Consequence Operator it is possible to determine if a given clause can be an abductive goal for a given logic program.

Doctor Of Philosophy - Phd, Cognitive Science

My thesis concerned modeling of abductive reasoning in a neuro-symbolic system which combines logic programs and artificial neural.

Master'S Degree, Cognitive Science

Master Of Science - Ms, Mechanics And Mechanical Engineering, Mechatronics

Master Of Science - Ms, Mechanical Engineering

Master Of Science - Ms, Mechanical Engineering

Bachelor Of Science - Bs, Mechanics And Mechanical Engineering