- LLMs: deployed and evaluated LLM models for semantic search, RAG and summarization of scientific documents with queries in natural language by the user, in a complete micro-services pipeline for data collection, pre-processing and answer generation.- AWS: created Lambda functions with/without using container images, deployed and maintained for improved remote sensing and geospatial data processing.- Data engineering: developed a remote sensing image/multispectral data pipeline for collecting, transforming and storing in MySQL databases and AWS S3 Buckets for improved visualization and localized area fast processing.- Software engineering: Containerization with Docker, API building with FastAPI, test-driven development for code quality, API authentication using Auth0, asynchronous processing and microservices integration with Kafka.- Data analysis and visualization on: > Agriculture: deployed ML models for analysis and prediction of hydroponically cultivated microgreens’ health and nutrient content, rain water flow prediction from Digital Elevation Maps for crop management and workspace safety. > Climate change: predicted future land degradation based on remote sensing data, land use/cover, soil element properties as part of a tool for Land Degradation Neutrality Initiative of United Nations Convention to Combat Desertification-UNCCD. > Food safety: performed tabular and image data analysis on design of plant-protein based products , quality assessment of fish products with ML models from IoT and spectral sensors.

- Experimented with various popular and state-of-the-art instance segmentation architectures (Mask R-CNN family) and backbone topologies (Swin, ConvNeXt among others), using Python and MMDetection, a Pytorch-based library.- Deployed an instance segmentation model using Docker and FastAPI, as an interactive component of my funding project's web platform.- Executed data pre-processing, cleaning and annotating to prepare datasets for supervised learning models using Python and CVAT.- Designed and conducted experiments in real-life and laboratory conditions for time-static and time-lapse image datasets creation.- Published 2 articles in peer-reviewed journals, presented 2 articles in peer-reviewed conferences and a poster.- During my Ph.D. I supervised: > A Master by Research (MScR) student running experiments for data collection, annotating images for dataset creation and applying statistical analysis for real-life crop size estimation. > A Research Assistant deploying DL detection architectures and developing a crop growth tracking algorithm on timelapse image data. > Two assistant annotators cleaning and annotating datasets for DL model supervised training.- Additional Modules selected: > Generative AI: Language Models Semester Project: Using python I created a Cryptocurrency Trading Assistant based on Assistants API, OpenAI. The pipeline included RAG on crypt-relevant documents/reports, API calls on websites with crypto trading information and web scrapping of relevant published articles. The LLM provided guidance and answers to user's queries by collecting up-to-date info on cryptocurrencies, defining trading/investment strategies based on expert information present in RAG documents and current state/sentiment/news. > Hyperspectral Imaging Learnt techniques for analyzing large 3D data cubes of hyperspectral data for various regression and classification tasks. > Scientific Writing Workshop

Implemented machine learning models on satellite data for climatic analysis.