Sayan Chakraborty

Dynamic Data Engineer with a proven track records, adept in Azure Databricks and Python. Excelled in building data pipelines and enhancing data-driven decisions with a 95% bug resolution rate. Skilled in Python, Machine Learning, and collaborative problem-solving, ensuring impactful business insights and system stability.

Detecting Aggression in Social Media Post using Transfer learning and Deep Learning methods, Applied text preprocessing techniques like tokenization and feature extraction., Established a baseline using Multinomial Naive Bayes., Advanced models included CNN and LSTM architectures., Fine-tuned BERT (LLM) for detecting aggression based on specific dataset., https://github.com/sayan936/Social-Media-Aggression-Detection Cricket Player Statistics Chatbot Development, Integrated Streamlit Chat for seamless messaging, enhancing user engagement and providing a conversational interface., Employed HuggingFacePipeline with the model 'google/flan-t5-base' for natural language understanding, enabling accurate responses to complex cricket-related queries., Configured FAISS (Facebook AI Similarity Search) for fast and efficient similarity search in large datasets, improving response times and relevance of chatbot answers., Created custom PromptTemplate for structured query handling, ensuring the chatbot provides relevant and contextually correct information to user inquiries., Developed a RetrievalQA chain integrating the chatbot with a retrieval system, allowing for accurate and data-driven answers based on T20I cricket player statistics., https://github.com/sayan936/cricbot LLama Model Fine-Tuning for Enhanced Language Understanding, Fine-tuned the LLama 2-7B language model on the OpenOrca dataset to improve natural language processing capabilities., Employed advanced NLP tools and techniques, including Transformers and BitsAndBytes, for efficient model optimization., Implemented PEFT (Parameter-Efficient Fine-Tuning) strategies to enhance model adaptation with minimal parameter adjustments., Utilized HuggingFace Hub for effective dataset management and versioning. Empathy Score Detection, Cleaned raw data and performed EDA using python libraries., Features Extraction like Pupil Diameter, Saccade Fixations., Time Series Analysis of movements of eye at different intervals., Using ML algorithms like Random Forest and Linear Regression., https://github.com/sayan936/Empathy/tree/main

German, A1

• Data Scientist Internship, BlackCoffer, 10/01/23, 02/29/24, Developed a Python web scraping script for data extraction needs, demonstrating expertise in Python and web scraping., Creating python script for organizing data for client promotional activity., Utilized advanced SQL queries for efficient data retrieval from databases, contributing to improved data-driven decision-making.
• AI/ML Development, Baavlibuch, 08/01/23, 10/31/23, Worked on the development of a chatbot for patient education., Modularised the code for semantic search on BioOntology portal., Deployed code using Django., Did syntactic string matching using ngrams and naïve methods.