I am a Machine Learning Engineer with ±3 years of experience in Computer Vision & Data Science. I have a Bachelor Degree in Computer Science and currently pursuing Master Degree in Artificial Intelligence. I had strong experience in: object detection, object classification, object tracking, semantic segmentation, face matching. named entity recognition, sentiment analysis, keyword extraction.

In my arsenal a lot of tools!

• Frameworks: PyTorch, TensorFlow, React.js, Lightning.
• Libraries: Pandas, Numpy, Scipy, Scikit-Learn, Albumentations, Torchvision, Torchmetrics, Ultralytics.
• Tools: LabelStudio, Roboflow, AWS EC2, AWS S3, GCP S3, DVC, Git.

1. DeepLabV3-CelebHQ. Semantic segmentation of a face based on a dataset CelebHQ with the use of DeepLabV3. Model successfully recognizes nose, eyes, ears, lips, face, and brows.
2. Efficient Graph-Based Segmentation. Highly efficient implementation of a classic segmentation algorithm that extracts elements based on its boundaries. Algorithm is put into a package to use out of the box. Can be used to generate a pre-train data for early R-CNN object detection model.
3. Cancer tissues classification. Comparison of ResNet, AlexNet, and VGG in predicting whether there is metastasis in a competition Histopathologic Cancer Detection.
4. AlexNet Implementation & Training. Trainined on a ImageNet fairly old classificator. It was rather at the start of my career.
5. Object Classification with ViT. Implemented a Visual Transformer architecture for classification of the hand-drawn shapes on a Drawn Shapes.

1. Named Entity Recognition with CRFSuite from Sklearn. Classification words in a sentence into a Location/Person/Misc from the WikiNER dataset. Achieved high F1 score.
2. News Scrapping. With use of Scrapy wrote a crawler that extracts specific news and puts it into the database for further use. Scrapped info includes: author, link, language, header, article body, published date.

[Under Construction]

1. Surveillance POC. A complex system connecting RPI Cameras with a server for online video translation. Server not only writes video to the disc, but also detects faces using YOLO and then uses ArcFace to compare features of the faces present on the camera with the faces present in the vector ChromaDB database to alert security about potential threats.
2. Santa Telegram Bot. Self-hosted telegram bot that has functionality to create a party & invite members. It also uses OpenAI API to help assist with the present based on a description the person you will give present to provided.
3. SimpleCOCO. Self-hosted tool for dataset labelling. It was created as a client for dataset generation tool for manual filtering of generated content. It can connect to the local server as well as to the cloud server, or server hosted on another machine.
4. Satellite Dataset Extraction. Tool for using API of Sentinel2 platform. It allows user to extract patches of a specific region for a specified period of time. It can accept multiple coordinates, timeranges, etc. Although requires Sentinel2 Token to operate.
5. Auto Subtitles and Audio Translation with Amazon Transcribe and Amazon Polly. Used AWS for deployment of a Streamlit application that enabled users transcribed videos (emerged when I got tired of searching for subtitles on Indian lectures, no offence, just indian accent is tough for me sometimes).