Hey, I'm Kosmas, an AI Engineer passionate about open-source NLP and deep learning projects. My current focus is on increasing the safety, efficiency, and usefulness of generative AI models through a variety of techniques.
About me
Iām a 25-year-old software engineer with a passion for Artificial Intelligence. My journey began at the Aristotle University of Thessaloniki, where I received my M.Eng. degree in Electrical and Computer Engineering. I am currently working as an Artificial Intelligence Engineer at Cogninn.
My favorite part of programming is the problem-solving aspect. I enjoy the sense of accomplishment of finally solving a problem. During my studies, I gained experience in Python, C/C++, SQL, and Matlab while working on several projects.
Currently, my main focus is on developing and improving generative AI models in the Natural Language Processing (NLP) area. I am primarily interested in text production, summarization, and image-to-text recognition. In addition, sometimes I work on computer vision projects, including object detection and image classification.
In my spare moments, I enjoy playing video games, playing football, and learning new skills. Specifically, in the last few months, I started learning video editing, and I can't wait to finish my next travel vlog. Until then, grab some popcorn and check out my most recent video from Iceland on my YouTube channel.
I created this website to share my thoughts and connect with people. Feel free to reach out to me on LinkedIn or via email.
My projects
Txtify
Open-source application that converts audio and video files into text using advanced AI models for fast and accurate transcriptions, available for self-hosting.
- Python
- Whisper
- Hugging Face
- DeepL
- FastAPI
JoyFill
Open-source web application to enhance sentences by providing positive suggestions using BERT for text completion and DistilBERT for sentiment analysis.
- Python
- BERT
- FastAPI
- Docker
- Hugging Face
Lee GOL CNN
This project combines Conway's Game of Life cellular automaton with a Convolutional Neural Network (CNN) to predict the shortest path in a matrix after a certain number of iterations.
- Python
- CNN
Neural Networks - Deep Learning
Digit recognition from 0ā9 using various models, including SVM, kNN, NCC, and Sequential Neural Networks with PCA.
- Python
- NN
- Deep Learning
- Digit Recognition
Non Local Means
Exercise in Parallel Programming: Implementing Non Local Means algorithm for image denoising using CUDA for significant speedup (up to 200 times).
- C
- CUDA
- Image Processing
My skills
- Python
- NLP
- LLMs
- GANs
- LSTM
- CNN
- RNN
- Git
- HuggingFace
- Docker
- VSCode
- NumPy
- TensorFlow
- PyTorch
- Pandas
- Matplotlib
- OpenCV
- sklearn
- Jupyter Notebook
- C
- C++
- SQL
- Java
- JavaScript
- Bash
- MATLAB
- LaTeX
- Javascript
- NodeJS
- React
- AutoML
- Explainable AI
- Neural Architecture Search
- Federated Learning
- Transfer Learning
- Active Learning
- FastAPI
- MongoDB
- SQLite
My experience
Embedded Systems Engineer Intern - NET2GRID
Thessaloniki, Greece
ā¢ Connected AWS Services (IoT Core, DynamoDB, API Gateway, Lambda) with IoT devices, enhancing system capabilities. ā¢ Developed and managed a Greengrass Group with Core and Edge devices, optimizing system performance. ā¢ Built MQTT bridge connections and implemented a Sequential Keras Model for accurate predictions, improving data communication and analysis.
Jul 2021 ā Sep 2021Embedded Systems Engineer - NET2GRID
Thessaloniki, Greece
ā¢ Collaborated with a 6-member team and developed an embedded firmware library for IoT device ESP32 using C.
Oct 2021 ā Jun 2022Freelancer
Remote,Greece
ā¢ Delivered Deep Learning projects using NLP, Computer Vision, and Data Analysis for various clients. ā¢ Instructed over 50 university students in AI, Maths, and Programming, achieving 95% positive feedback and significantly enhancing student performance.
Jan 2023 ā Sep 2023
AI Engineer - Cogninn
Larissa, Greece
ā¢ Led the development of AutoDL, a web platform managing diverse data inputs (images, text, videos, tabular data), and architected ML/DL pipelines, resulting in increased model training accuracy and significant business impact. ā¢ Implemented complex image processing tasks(classification, object detection, segmentation, OCR) and text analysis functionalities(classification, summarization). ā¢ Engineered video object detection and tabular regression capabilities, increasing processing speed by 15%. ā¢ Deployed RESTful API endpoints using FastAPI and integrated MongoDB and SQLite for effective data storage.
Sep 2023 ā Aug 2024Contact me
Please contact me directly at louiskmeta@gmail.com or through this form.