Worked with unstructured text data on AWS, enhancing NLP techniques with Pandas to boost performance by ~70% and brought transformer based model and improved system performance 🚀:1. Customized Regular Expressions for crucial data extraction 🧩2. Conducted Text Preprocessing 📝3. Utilized Topic Modeling 📊4. Employed Word Embeddings (TF-IDF, FastText, Word2Vec) 📖5. Applied ML Models (SVM, Naive Bayes, Decision Tree) 🤖6. Implemented Data Balancing via SMOTE ⚖️7. Fine-tuned Document AI multimodel 📄8. Utilized AWS services (ECS, Lambda, Step Function, Terraform, CloudFormation) ☁️9. Created ETL process to load data from PROD, transformed and load for QuickSight dashboard to analyze user's usage for business. 📊📈10. Proficient in Transformer-based model fine-tuning with Huggingface 🤗.11. Effectively train and deploy transformer based model using Sagemaker.Managed large data volumes with S3 and Sagemaker while optimizing performance with PySpark. 📦💡

Project was done under PIVP Invest Nova Scotia grant. - Led and executed a production-ready project utilizing LLM (Large Language Model).- Proposed and implemented solutions, utilizing the RAG process for decision-making.- Conducted bi-weekly work presentations to ensure alignment with business requirements.Project Highlights 🌟:Data Extraction 📊: - Overcame challenges with web PDF and webpage data sources. - Leveraged llmsherpa and langchain libraries for effective data chunking. Vector Database 🗄️: - Utilized Pinecone for efficient vector data storage.Embedding Model 🤖: - Compared and visualized multiple models for optimal performance.Dataset Creation 📚: - Developed context-question pairs for embedding fine-tuning. - Utilized the 7B open-source model, refining prompts for high-quality datasets.Achievements 🏆: - Achieved an impressive 83% hit rate with fine-tuned datasets. - Improved hit rates by ~10% compared to the best open-source models.Generative Model 🔄: - Extracted relevant texts from the vector database for LLM input. - Enhanced performance through effective prompt engineering.

Check Data mining Assignment and Check undergrad student submission.

Working with multiple data sources to get the required data. Created process; proposed calculations based on feature requests and fit that work in end to end workflow.

Role: Research Associate🌱🔍 Problem Analysis & Solutions:- Diligently analyzed issues and devised effective solutions.- Focused on extracting carbon emissions data from diverse documents.🌐 Information Extraction:- Utilized Natural Language Processing for Information Extraction.- Web scraped online sources and extracted data from PDFs with Apache TIKA.📋 Text Data Refinement:- Systematically organized and refined textual data.- Employed NLTK for predefined entity extraction using grammar rules.💼 Business Enhancement:- Augmented the database with user-friendly flags and source links.- Streamlined data access and analysis, aiding efficient decision-making. 🚀

Role: Senior Technical Developer 👨‍💻1. Efficient Project Management 🚀: Managed multiple projects simultaneously, prioritizing tasks effectively while adhering to Agile methodologies. Quickly adapted to new tech like Blue Prism, providing valuable support.2. Data Extraction and Transformation 📊: Extracted iMatch data using complex SQL in DataIKU, applying transformations with "pandas." Visualized workflows in Mural and presented sprint progress regularly.3. Data Analysis and Visualization 📈: Conducted in-depth database analysis, sourced data, and connected results to Tableau for visualization.4. RPA BOT Management and Support 🤖: Resolved issues, enhanced RPA BOTs, and facilitated team communication. Contributed to documentation and addressed server issues promptly.5. Data Processing for Machine Learning 🧠: Retrieved data from various sources, applied business logic, and prepared data for Machine Learning models within the Kanban framework. 🔄

Subject: Data MiningTasks: Resolve student's queries, Assess Assignment and provide feedback.

Python - Fetched the data from the Mysql and validate data. Add those data into Excel and add some comments based on received data. Agile methodology - followed Scrum framework in the assigned project.Optimization/ Performance - Added multithreading with multiprocessing. Cutover redundant steps in the process to make it fast.