Worked on the 'Locomotive-AI' project, utilizing LiDAR and 2D imagery for locomotive detection and track analysis. Developed algorithms to detect locomotives on the main track, identify turnaround points, segment the track, and trigger alerts as necessary. Main technical focus on Python-based AI and computer vision.

Executed a straightforward Image Processing project for real-time vehicle detection and classification from live camera streams. Implemented functions for vehicle counting and peak-hour vehicle analysis. Tasked with identifying and categorizing vehicles into five distinct classes, such as red taxi, red bus, and more. Developed the project within the Django framework. Leveraged AWS cloud services for hosting and deployment. Utilized the Python programming language for image processing and classification tasks.

Collaborated with a prominent big data scraping firm, specializing in frequent data extraction from major client websites. Proficiently employed a tech stack centered around Python, including Scrapy, an in-house API for handling captchas and bans, Beautiful Soup, HTML, and regular expressions (regex). Addressed challenges related to website bans by implementing proactive measures, such as rotating proxies and IP management, ensuring uninterrupted data scraping operations. Processed and delivered scraped data for utilization in AI algorithms, contributing to valuable insights and informed decision-making. Developed a Selenium-based web scraper specifically for extracting job listings from LinkedIn, followed by data analysis. Applied question-answering algorithms to extract seniority levels, salary information, and other relevant details from the scraped job data.

Received hotel vendor invoices written in Danish, a major European language, for processing. Conducted Name Entity Recognition (NER) and data analysis on these invoices, which were provided in PDF and image formats. Overcame the significant challenge of dealing with inconsistent and non-standard data formats. Designed a robust solution, involving the training of a computer vision algorithm, YOLOv8, to identify relevant areas on the invoices. Utilized Optical Character Recognition (OCR) to extract text from the identified areas accurately. Trained BERT (Bidirectional Encoder Representations from Transformers) for Named Entity Recognition (NER) to categorize and extract specific entities from the extracted text. Designed the architectural framework of the product, incorporating it into a Flask-based API for seamless integration and scalability. Managed a team of 5 individuals, overseeing their roles and responsibilities in the project. Established direct communication channels with clients to ensure project alignment and meet their specific requirements.

Employed computer vision algorithms to train models for product recognition on supermarket shelves. Implemented image stitching techniques to create comprehensive views of supermarket shelves for improved analysis. Designed the architecture of the product, incorporating microservices for scalability and modularity. Utilized Minio as a storage solution to store machine learning models and associated data. Leveraged Docker for containerization, ensuring consistency and portability of the deployment environment. Acquired proficiency in FastAPI and utilized it to wrap the trained models, making them accessible via API endpoints. Managed data storage using PostgreSQL, ensuring efficient and secure data management within the project.

Developed matching algorithms to identify relevant connections and associations within the scraped data. Trained a classifier to categorize creators into their primary niches based on their content and profiles. Collaborated with a team to design and build an analytics dashboard for data visualization and insights. Utilized GPT-3 to generate biographies for Instagram profiles, enhancing profile descriptions. Leveraged Python for Natural Language Processing (NLP) and Computer Vision tasks. Employed ElasticSearch and Kibana for data indexing, searching, and visualization within the project.